research paper on semantic web services

An official website of the United States government

The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

• Publications
• Account settings

Preview improvements coming to the PMC website in October 2024. Learn More or Try it out now .

• Advanced Search
• Journal List
• Int J Clin Pract
• v.2022; 2022

Semantic Web in Healthcare: A Systematic Literature Review of Application, Research Gap, and Future Research Avenues

A. k. m. bahalul haque.

1 Electrical and Computer Engineering, North South University, Dhaka 1229, Bangladesh

B. M. Arifuzzaman

Sayed abu noman siddik, tabassum sadia shahjahan, t. s. saleena.

2 PG & Research Department of Computer Science, Sullamussalam Science College Areekode, Malappuram, Kerala 673639, India

Morshed Alam

3 Institute of Education and Research, Jagannath University, Dhaka 1100, Bangladesh

Md. Rabiul Islam

4 Department of Pharmacy, University of Asia Pacific, 74/A Green Road, Farmgate, Dhaka 1205, Bangladesh

Foyez Ahmmed

5 Department of Statistics, Comilla University, Kotbari, Cumilla, Bangladesh

Md. Jamal Hossain

6 Department of Pharmacy, State University of Bangladesh, 77 Satmasjid Road, Dhanmondi, Dhaka 1205, Bangladesh

Associated Data

The data used to support the findings of this study are included within the article.

Today, healthcare has become one of the largest and most fast-paced industries due to the rapid development of digital healthcare technologies. The fundamental thing to enhance healthcare services is communicating and linking massive volumes of available healthcare data. However, the key challenge in reaching this ambitious goal is letting the information exchange across heterogeneous sources and methods as well as establishing efficient tools and techniques. Semantic Web (SW) technology can help to tackle these problems. They can enhance knowledge exchange, information management, data interoperability, and decision support in healthcare systems. They can also be utilized to create various e-healthcare systems that aid medical practitioners in making decisions and provide patients with crucial medical information and automated hospital services. This systematic literature review (SLR) on SW in healthcare systems aims to assess and critique previous findings while adhering to appropriate research procedures. We looked at 65 papers and came up with five themes: e-service, disease, information management, frontier technology, and regulatory conditions. In each thematic research area, we presented the contributions of previous literature. We emphasized the topic by responding to five specific research questions. We have finished the SLR study by identifying research gaps and establishing future research goals that will help to minimize the difficulty of adopting SW in healthcare systems and provide new approaches for SW-based medical systems' progress.

1. Introduction

The detection and remedy of illnesses through medical professionals are expressed as healthcare. The healthcare system consists of medical practitioners, researchers, and technologists that work together to provide affordable and quality healthcare services. They tend to generate considerable amounts of data from heterogeneous sources to enhance diagnostic accuracy, elevate quick treatment decisions, and pave the way for the effective distribution of information between medical practitioners and patients. However, it is necessary to organize those valuable data appropriately so that they can fetch those, while required.

One of the main challenges in utilizing medical healthcare data is extracting knowledge from heterogeneous data sources. The interoperability of well-being and clinical information poses tremendous obstacles due to data irregularity and inconsistency in structure and organization [ 1 , 2 ]. This is also because data are stored in various authoritative areas, making it challenging to retrieve knowledge and authorize a primary route along with information analysis. The information from a hospital can prove to be very useful in healthcare if these data are shared, analyzed, integrated, and managed regularly. Again, platforms that provide healthcare services also face dilemmas in automating time-efficient and low-cost web service arrangements [ 3 ]. This indicates that meaningful healthcare solutions must be proposed and implemented to provide extensive functionality based on electronic health record (EHR) workflows and data flow to enable scalable and interoperable systems [ 4 ], such as a blockchain-based smart e-health system that provides patients with an easy-to-access electronic health record system through a distributed ledger containing records of all occurrences [ 5 – 8 ]. A standard-based and scalable semantic interoperability framework is required to integrate patient care and clinical research domains [ 9 ]. The increasing number of knowledge grounds, heterogeneity of schema representation, and lack of conceptual description make the processing of these knowledge bases complicated. Non-experts find mixing knowledge with patient databases challenging to facilitate data sharing [ 10 ]. Similarly, ensuring the certainty of disease diagnosis also becomes a more significant challenge for health providers. Brashers et al. [ 11 ] in their work examined the significance of credible authority and the level of confidence HIV patients have in their medical professionals. Many participants agreed that doctors might not be fully informed of their ailment, but they emphasized the value of a strong patient-physician bond. With the help of big data management techniques, these challenges can be minimized. Likewise, Crowd HEALTH aims to establish a new paradigm of holistic health records (HHRs) that incorporate all factors defining health status by facilitating individual illness prevention and health promotion through the provision of collective knowledge and intelligence [ 11 – 13 ]. Another similar approach is adopted by the beHealthier platform which constructs health policies out of collective knowledge by using a newly proposed type of electronic health records (i.e., eXtended Health Records (XHRs)) and analysis of ingested healthcare data [ 14 ]. Making healthcare decisions during the diagnosis of a disease is a complex undertaking. Clinicians combine their subjectivity with experimental and research artifacts to make diagnostic decisions [ 9 ].

In recent years, Web 2.0 technologies have significantly changed the healthcare domain. However, in proportion to the growing trend of being able to access data from anywhere, which is primarily driven by the widespread use of smartphones, computers, and cloud applications, it is no longer sufficient. To address such challenges, Semantic Web Technologies have been adopted over time to facilitate efficient sharing of medical knowledge and establish a unified healthcare system. Tim Berners-Lee, also known as the father of the web, first introduced Semantic Web (SW) in 1989 [ 15 ]. The term “Semantic Web” refers to linked data formed by combining information with intelligent content. SW is an extension of the World Wide Web (WWW) and provides technologies for human agents and machines to understand web page contents, metadata, and other information objects. It also provides a framework for any kind of content, such as web pages, text documents, videos, speech files, and so on. The linked data comprise technologies such as Resource Description Framework (RDF), Web Ontology Language (OWL), SPARQL, and SKOS. It aims to create an intelligent, flexible, and personalized environment that influences various sectors and professions, including the healthcare system.

Data interoperability can only be improved when the semantics of the content are well defined across heterogeneous data sources. Ontology is one of the semantic tools, which is frequently used to support interoperability and communication between software, communities, and healthcare organizations [ 16 , 17 ]. It is also commonly used to personalize a patient's environment. Kumari et al. [ 18 ] and Haque et al. [ 19 ] proposed an Android-based personalized healthcare monitoring and appointment application that considers the health parameters such as body temperature, blood pressure, and so on to keep track of the patient's health and provide in-home medical services. Some existing ontologies of medicine are Gene, NCI, GALEN,LinkBase, and UMLS [ 20 ]. They have also been used in offering e-healthcare systems based on GPS tracking and user queries. Osama et al. proposed two ontologies for a medical differential diagnosis: disease symptom ontology (DSO) and patient ontology (PO) [ 21 ]. Sreekanth et al. used semantic interoperability to propose an application that brings together different actors in the health insurance sector [ 22 ]. Semantic Web not only enables information system interoperability but also addresses some of the most challenging issues with automated healthcare web service settings. SW combined with AI, IoT, and other technologies has produced a smart healthcare system that enables the standardization and depiction of medical data [ 1 , 23 , 24 ]. In terms of economic efficiency, the Semantic Web-Based Healthcare Framework (SWBHF) is said to benchmark the existing BioMedLib Search Engine [ 25 ]. SW also offered a new user-oriented dataset information resource (DIR) to boost dataset knowledge and health informatics [ 26 ]. This technology is also used in the rigorous registration process to discover, classify, and composite web services for the service owner [ 4 ]. To provide answers to medical questions, it has been integrated with NLP to create RDF datasets and attach them with source text [ 27 ]. Babylon Health, which enables doctors to prescribe medications to patients using mobile applications, has benefited from the spread of semantic technology. Archetypes, ontology, and datasets have been used in web-based methods for diagnosing colorectal cancer screening. Clinical information and knowledge about disease diagnosis are encoded for decision making with the use of ontological understanding and probabilistic reasoning. The integration of pharmaceutical and medical knowledge, as well as IoT-enabled smart cities, has made extensive use of SW technologies [ 8 ]. To put it briefly, this emerging technology has revolutionized the healthcare and medical system.

Despite its relevance, researchers who looked into the benefits of SW efforts showed substantial deficiencies in the wide range of semantic information in the medical and healthcare sectors. To the best of our knowledge, no previous systematic literature review (SLR) has been published on the Semantic Web and none of the research has previously classified the precise application area in which SW can be applied. Furthermore, there was an absence of research questions in the previous literature for analyzing and comparing similar works in order to understand their flaws, strengths, and problems.

In this study, we present a systematic review of the literature on Semantic Web in healthcare, with an emphasis on its application domain. It is absolutely essential to point the SW user community in the right direction for future research, to broaden knowledge on research topics, and to determine which domains of study are essential and must be performed. Thus, the current SLR can help researchers by addressing a number of factors that either limit or encourage medical and healthcare industries to employ Semantic Web technologies. Furthermore, the study also identifies various gaps in the existing literature and suggests future research directions to help resolve them. The research questions (RQs) that this systematic review will seek to answer are as follows. ( RQ1 ) What is the research profile of existing literature on the Semantic Web in the healthcare context? ( RQ2 ) What are the primary objectives of using the Semantic Web, and what are the major areas of medical and healthcare where Semantic Web technologies are adopted? ( RQ3 ) Which Semantic Web technologies are used in the literature, and what are the familiar technologies considered by each solution? ( RQ4 ) What are the evaluating procedures used to assess the efficiency of each solution? ( RQ5 ) What are the research gaps and limitations of the prior literature, and what future research avenues can be derived to advance Web 3.0 or Semantic Web technology in medical and healthcare?

This research contributes in a number of ways. This paper's main focus is centered on the collection of some statistical data and analysis results that are mostly focused on the adoption of SW technologies in the medical and healthcare fields. First, we gathered data from five publishers, including Scopus, IEEE Xplore Digital Library, ACM Digital Library, and Semantic Scholar, to thoroughly review, analyze, and synthesize past research findings. Furthermore, the current study does not focus on a specific theme, rather, it offers a broad overview of all possible research themes related to the use of SW in healthcare. Finally, this SLR identifies gaps in the existing literature and suggests a future research agenda. The primary contributions of our study are listed as follows:

• To find out the up-to-date research progress of SW technology in medical and healthcare.
• To open up new technical fields in healthcare where SW technologies can be used.
• To identify all the constraints in the healthcare industry during the adoption of SW technologies.
• To identify key future trends for semantics in the healthcare sector.
• To analyze and investigate alternative strategies for ensuring semantic interoperability in the healthcare contexts.

This review paper is organized as follows. Section 1 introduces Semantic Web technologies in healthcare followed by Section 2 which describes the methodology followed, the inclusion/exclusion criteria, and the data extracted and analyzed in this literature review paper. Section 3 elaborately discusses different thematic areas, and Section 4 presents the research gaps to address future research agendas. Section 5 presents a detailed discussion of the specified RQs. Lastly, Section 6 consists of the conclusion for this SLR.

2. Methodology

A systematic review is a research study that looks at many publications to answer a specific research topic. This study follows such a review to examine previous research studies that include identifying, analyzing, and interpreting all accessible information relevant to the recent progress of pertinent literature on Web 3.0 or Semantic Web in medical and healthcare or our phenomenon of interest. In the advancement of medical and healthcare analysis, numerous SLRs have been undertaken with inductive methodologies to identify major themes where Semantic Web technologies are being adopted [ 28 , 29 ]. In our study, we adopted the procedures outlined by Keele with a few important distinctions to assure the study's transferability, dependability, and transparency, emphasizing and documenting the selection method [ 30 ]. The guidelines outlined in that paper were derived from three existing approaches used by medical researchers, two books written by social science researchers, and a discussion with other academics interested in evidence-based practice [ 8 , 31 – 40 ]. The guidelines have been modified to include medical policies in order to address the unique challenges of software engineering research.

Our study sequentially conducted an SLR to accomplish the precise objectives. At first, we planned the necessary approach to identify the problems. Next, we collected related study materials and retrieved data from them. Finally, we documented the findings and carried out the research in the following steps (see Figure 1 ) maintaining its replicability as well as precision.

• Step 1 . Plan the review by finding appropriate research measures to detect corresponding documents.
• Step 2 . Collect analyses by outlining the inclusion and exclusion criteria to assess their applicability.
• Step 3 . Extract relevant data using numerous screening approaches to use accordingly.
• Step 4 . Document the research findings.

SLR methodology and protocols.

2.1. Planning the Review

The very first stage in conducting SLR is to identify the needs for a specific systematic review, outline the research questions, design a procedural review, and offer a study framework to assist the investigation in subsequent phases to identify the systematic review's significant objectives. This phase begins with the identification of needs for the proposed systematic review. Section 1 of this paper went into detail about why a systematic review of Semantic Web technologies in healthcare was deemed necessary. Following that, the definition of research questions, the selection of a synthesis method, initial keywords, and databases are given. To begin, we devised the RQs for this SLR in order to gain a comprehensive understanding of the semantic-based solutions in the field of healthcare. Defining research questions is an important part of conducting a systematic review because they guide the overall review methodology. Based on the objectives, we conducted a pilot study of a systematic review of fifteen sample studies, resulting in the broad application of the Semantic Web to a specific niche, refinement of research questions, and redefinition of the review research protocol. To find relevant scientific contributions for our RQs, we used Scopus, IEEE Xplore Digital Library, ACM Digital Library, and Semantic Scholar. Furthermore, we utilized the primary term “Web 3.0 or Semantic Web” to search the databases and then identified and refined the comprehensive keywords that would be used as search strings. We did not limit our search to a single period instead; we looked at all linked studies.

2.2. Collecting Analyses

A systematic review's unit of analysis is crucial since it broadens the scope of the overall approach. This study aims to better understand how Web 3.0 or Semantic Web technologies are employed in medical and healthcare settings, as well as to identify the extent to which they have been applied. We have selected academic research articles and journals as the unit of analysis for our SLR. We specified inclusion and exclusion criteria to narrow the investigation in the following study selection process, as shown in Table 1 . To gather our search phrases, we used a nine-step procedure as mentioned in [ 41 ]. The studies obtained from online repositories were compared with exclusion criteria to select peer-reviewed papers and eliminate any non-peer-reviewed studies. To perform this review, we employed decisive exclusion criteria to identify grey literature, which included white papers, theses, project reports, and working papers. To remove language barriers, we only selected papers written in English. We did not consider any review papers or project reports to maintain the quality. Older publications that have never been cited were excluded from the review to explore the potential value of Web 3.0 and SW technologies in medical and healthcare.

Criteria for inclusion and exclusion.

2.3. Extracting Relevant Data

Initially, we searched for papers in Google Scholar with “Web 3.0 in medical and healthcare” keywords. However, reviewing the title and abstract from the top 50 articles further improved the search keyword to develop a more appropriate search string. The top search string (“Semantic Web” OR “Web 3.0”) AND (“Healthcare” OR “medical”) was used in Scopus, IEEE Xplore Digital Library, ACM Digital Library, and Semantic Scholar to find related papers for our SLR on 22 January 2022. We found a total of 4137 papers, including 2237 from IEEE Xplore Digital Library, 1761 from Scopus, 103 from Semantic Scholar, and 36 from ACM Digital library. Primary review grasped articles up to 2001. So, all the identified publications were from 2001 to 2021. Four authors performed the screening method through different stages. After each step, a discussion session was held to finalize the step and move further.

At first, we checked for any duplicate articles from both indexing databases. We eliminated available duplicate articles by checking the Digital Object Identifier (DOI) and the research heading. After removing the duplicate articles, we were left with 1923 articles. After that, titles, keywords, and abstracts were read as part of the preliminary screening process. During the screening procedure, articles were divided into three categories: retain, exclude, and suspect. After removing articles unrelated to Web 3.0 or Semantic Web in medical and healthcare, only 1741 articles were retrained. Upon analyzing the contents of both suspect and retain studies using the inclusion and exclusion criteria listed in Table 1 , we were left with 343 publications. Following that, we read the full text of the articles that were picked, and we were left with 54 papers being considered for our conclusive stage. Finally, we applied the snowballing strategy, also known as the citation chaining technique [ 42 ]. Surprisingly, this step resulted in the addition of another ten studies (7 from backward citation and three from forwarding citation). The final review pool thus comprised 65 papers being considered for our conclusive stage ( Figure 2 depicts the study selection process in detail).

Study selection process.

2.4. Document Research Findings

The shortlisted research papers were profiled using descriptive statistics, which include publication year, methodology, and publication sources [ 23 , 43 , 44 ]. According to the chronology of the number of publications, the majority of the research articles were published in 2013. However, between 2018 and 2021, the number declined. Figure 3 depicts the yearly (between 2001 and 2022) distribution of published papers.

The majority of the studies presented a framework for developing a medical data information management system. Web 3.0 technologies appear to be in their early phases of adoption, with scholars only recently becoming interested in the topic. A few other papers discussed medical data interchange mechanisms, diseases, frontier technology such as AI and NLP, and regulatory conditions. Nearly half of the research ( n  = 39) was published between 2001 and 2012, with the remaining studies ( n  = 26) published after that (see Figure 3 ). The Semantic Web theory gained widespread interest after the architect of the World Wide Web, Tim Berners-Lee, James Hendler, and Ora Lassila popularized it in a Scientific American article in May 2001 [ 15 ]. This trend also gained momentum in recent years, with John Markoff coining the term Web 3.0 in 2006 and Gavin Wood, Ethereum's co-founder, coining the word later in 2014.

Number of articles published yearly.

Medical and healthcare writings have been published in several renowned conferences, journals, book series, and events. The 65 shortlisted papers are distributed throughout 27 conference proceedings, 21 journals, and 17 book series. The descriptive analysis depicts that 65 shortlisted analyses were authored by 25 publishers, accompanied by Springer ( n  = 17), IEEE Xplore ( n  = 15), IOS Press ( n  = 6), ACM ( n  = 5), and Elsevier ( n  = 3). Only a few publishers published many studies. The reset included 15 publishers, each of whom only published one study. However, the majority of the papers were published in Lecture Notes in Computer Science (LNCS), CEUR Workshop Proceedings, and Studies in Health Technology and Informatics Series (see Figure 4 ). Furthermore, our SLR demonstrates the wide geographic span of existing research papers. The United States (11 articles), France (23 articles), India (9 articles), Canada (8 articles), Belgium (4 articles), and South Korea (4 articles) all had a significant number of studies. Figure 5 summarizes the past literature's geographical distribution.

Publication-source-wise distribution.

Country-wise article distribution.

According to the systematic literature review, the application of Semantic Web technologies in the field of healthcare is a prominent classical research theme, with many innovative and promising research topics. The number of Semantic Web publications and interest in healthcare has increased rapidly in recent years, and Semantic Web methods, tools, and languages are being used to solve the complex problems that today's healthcare industries face. Semantic Web technology allows comprehensive knowledge management and sharing, as well as semantic interoperability across application, enterprise, and community boundaries. This makes the Semantic Web a viable option for improving healthcare services by improving tasks such as standards and interoperable rich semantic metadata for complex systems, representing patient records, investigating the integration of Internet of Things and artificial computational methods in disease identification, and outlining SW-based security. While there are interesting possibilities for the application of Semantic Web technologies in the healthcare setting, some limitations may explain why those possibilities are less apparent. We believe one reason is a lack of support for developers and researchers. Semantic Web-based healthcare applications should be viewed as independent research prototypes that must be implemented in real-world scenarios rather than as a widget that is integrated with the Web 2.0-based solution. This study discusses the findings and future directions from two different perspectives. First, consider the potential applications of Semantic Web technologies in different healthcare scenarios and also look at the barriers to their practical application and how to overcome them (see Section 3 ). Last, the fourth (see Section 4 ) section discusses the scope of research in Semantic Web-enabled healthcare.

3. Analysis of the Selected Articles: Thematic Areas

This section focuses on three key steps: summarizing, comparing, and discussing the shortlisted papers to describe and categorize them into common themes. To systematically analyze all 65 studies, we adopted the technique used in recently published SLRs [ 23 , 43 ]. After identifying and selecting relevant papers that could answer our research questions, we used the content analysis technique to classify, code, and synthesize the findings of those studies. A three-step approach was proposed by Erika Hayes et al., which was used to interpret unambiguous and unbiased meaning from the content of text data [ 45 ]. The steps were as follows: (a) the authors assigned categories to each study and a coding scheme created directly and inductively from raw data using valid reasoning and interpretation; (b) the authors immersed themselves in the material and allowed themes to arise from the data to validate or extend categories and coding schemes using directed content analysis; (c) the authors used summative content analysis, which begins with manifesting content and then expands to identify hidden meanings and themes in the research areas.

This thematic analysis answers the second research question (RQ2), “What are the primary objectives of using the Semantic Web, and what are the major areas of medical and healthcare where Semantic Web technologies are adopted?”, and this analysis architecture highlights five broad medical and healthcare-related research themes based on their primary contribution (see Table 2 ), notably e-healthcare service, diseases, information management, frontier technology, and regulatory conditions.

Derived themes and their descriptions.

Two themes, namely, IoT and cloud computing, were nevertheless left out since they lack a wide description that would be useful in developing a meaningful theme. Some of the papers from which we defined these two thematic areas were included in the selected themes based on their similarity to the chosen thematic areas. Figure 6 illustrates this categorization, with different themes' description, which emerged from our review.

Thematic description of Semantic Web approaches in medical and healthcare.

3.1. E-Healthcare Service

The use of various technologies to provide healthcare support is known as e-service in healthcare or e-healthcare service. While staying at home, a person can obtain all the necessary medical information as well as a variety of healthcare services such as disease reasoning, medication, and recommendation through e-healthcare services. It is similar to a door-to-door service. The Semantic Web or Web 3.0 plays a critical role in this regard. The Semantic Web offers a variety of technologies, including semantic interoperability, semantic reasoning, and morphological variation that can be used to create a variety of frameworks that improve e-healthcare services.

SW makes the task of sharing medical information among healthcare experts more efficient and easier [ 2 , 46 – 48 ]. A dataset information resource for medical knowledge makes the work more trouble-free and faster. A healthcare dataset information resource has been created along with a question-answering module related to health information [ 26 ]. Combining different databases can be more effective as it expands the information range of knowledge. In this respect, Barisevičius et al. [ 49 ] designed a medical linked data graph that combines different medical databases and they also developed a chatbot using NLP-based knowledge extraction that provides healthcare services by supplying knowledge about various medical information. Besides information sharing and database combining, Semantic Web-based frameworks can provide virtual medical and hospital-based services. A system has been created that provides medical health planning according to patient's information [ 50 ]. Concerning this, it could be very helpful if there is a system that can match patient requirements with the services. Such a matchmaking system has been developed to match the web services with the patient's requirements for medical appointments [ 51 ]. To provide hospital-based services, a Semantic Web-based dynamic healthcare system was developed using ontologies [ 17 ]. Disease reasoning is a vital task for e-healthcare services. A number of frameworks were developed that are used for reasoning diseases [ 49 , 52 , 53 ]. In addition, some authors implemented systems that provide support for sequential decision making [ 54 – 57 ]. Moreover, Mohammed and Benlamri [ 21 ] designed a system that could help to prescribe differential diagnosis recommendations. Grouping similar diagnosis patients can be useful to enhance the medication process. In this regard, Fernández-Breis et al. [ 58 ] created a framework to group the patients by identifying patient cohorts. Moreover, Kiourtis et al. [ 59 ] proposed a new device-to-device (D2D) protocol for short-distance health data exchange between a healthcare professional and a citizen utilizing a sequence of Bluetooth communications. Supplying medical information to people is one of the main tasks of e-healthcare services [ 58 ]. Before proceeding with a medical diagnosis, we need to be sure about the correctness of the procedure. Andreasik et al. [ 60 ] developed a Semantic Web-based framework to determine the correctness of medical procedures. Various systems for medical education were developed using Semantic Web technologies such as a web service delivery system [ 4 ], a web service searching system [ 61 ], and an e-learning framework for the patients to learn about different medical information [ 62 , 63 ]. Some articles discussed the rule-based approaches for the advancement of medical applications [ 64 , 65 ]. Quality assurance of Semantic Web services is necessary, and so a framework was created using a Semantic Web-based replacement policy to assure the quality of a set of services and replace it with a newly defined subset of services when the existing one fails in execution [ 3 ]. A framework was designed for Semantic Web-based data representation [ 66 ]. Meilender et al. [ 67 ] described the migration of Web 2.0 data into Semantic Web data for the ease of further advancement in Web 3.0.

Researchers used different Semantic Web services to convert the relational database to create Resource Description Framework (RDF) and Web Ontology Language (OWL)-based ontologies. It is done by extracting the instances from the relational databases and representing them into RDF datasets [ 21 , 55 , 57 , 62 ]. In some prior literature, many RDF datasets were created using Apache JENA 4.0 [ 4 ], different versions of protégé were used to construct and represent various healthcare ontologies [ 2 , 17 ], Apache Jena framework was used for OWL reasoning on the RDF datasets [ 50 , 53 ], and the EYE engine was used for reasoning [ 54 ]. Besides, Kiourtis et al. [ 68 ] developed a technique for converting healthcare data into its equivalent HL7 FHIR structure, which principally corresponds to the most used data structures for describing healthcare information. Furthermore, a sublanguage of F-logic named Frame Logic for Semantic Web Services (FLOG4SWS) and web services along with some features of Flora-2 was used to represent the ontology [ 51 ]. The authors of some papers used RDF and OWL for data representation of different ontologies [ 50 , 52 , 54 , 66 ]. Mohammed and Benlamri [ 21 ] offered a number of Semantic Web strategies for ontology alignment, such as ontology matching and ontology linking, and some used ontology mapping for the ontology alignment [ 58 , 66 ]. By combining RDF and semantic mapping features, Perumal et al. [ 69 ] provided a translation mechanism for healthcare event data along with Semantic Web Services and decision making. In addition, a linked data graph (LDG) is utilized to combine numerous publicly available medical data sources using RDF converters [ 49 ]. The works in [ 52 , 54 ] used Notation3 for data mapping. SPARQL was used as the query language for the database [ 2 , 17 , 50 , 52 , 57 ]. Besides, the Jena API was also used as a query language [ 21 ]. The Semantic Web's rules and logic were expressed in terms of OWL concepts using the Semantic Web Rule Language (SWRL) [ 55 , 57 ]. TopBraid Composer is used as the Semantic Web modeling tool [ 60 ].

There was no proof that the system created using semantic networks was able to share knowledge among healthcare services [ 2 , 46 , 48 ]. Researchers did not mention how a system can be integrated with different types of datasets in the world [ 2 , 47 ]. In their paper, Ramasamy et al. [ 3 ] did not mention whether the system could replace all types of services or not. Shi et al. [ 26 ] did not discuss the success rate of the datasets in their dataset information resource and the accuracies of different systems created with these datasets. No proper evaluation techniques have been given for linked data graph [ 49 ], Semantic Web service delivery systems [ 4 , 50 ], and Semantic Web reasoning system [ 52 , 53 ] in their studies. There is no discussion of the reliability and validity of numerous decision making and recommendation systems [ 21 , 54 , 70 ]. Podgorelec and Gradišnik [ 64 ] did not provide information about the betterment of the combined Semantic Web technologies and rule-based systems against other alternatives. Most of the articles discussed or offered various techniques to build different healthcare services, but there are only a few articles that implemented the proposed systems and tested them in a real-life context.

3.2. Diseases

This thematic area aims to specifically identify and discuss the contributions of Semantic Web technologies to reach interoperability of information in the healthcare sector and aid in the initial detection and nursing of diseases, such as diabetes, chronic conditions, cardiovascular disease, dementia, and so on. SW provides a framework to integrate medical knowledge and data for effective diagnosis and clinical service. They help to select patients, recognize drug effects, and analyze results by using electronic health data from numerous sources. The queryMed packages were proposed for pharmaco-epidemiologists that link medical and pharmacological knowledge with electronic health records [ 10 ]. This application searches for people with critical limb ischemia (CLI) with at least one medication or none at all and gives them healthcare recommendations. SW also emphasizes the study of phenotypes and their influence on personal genomics. The Mayo Clinic's project, Linked Clinical Data (LCD), facilitates the use of SW and makes it easier to extract and express phenotypes from electronic medical records [ 71 ]. It also emphasizes the use of semantic reasoning for the identification of cardiovascular diseases. Besides this, it aims to improve healthcare service quality for people suffering from chronic conditions. Proper planning and management are required for the better treatment and management of chronic diseases. Thus, the Chronic Care Model (CCM) provides knowledge-based acquisition to patients [ 72 ].

Ontology-based applications such as the Concept Unique Identifier (CUI) from Unified Medical Language System, Drug Indication Database (DID), and Drug Interaction Knowledge Base (DIKB) are widely used in the medical domain to establish mappings between medical terms [ 10 ]. In the context of ontology, the ECOIN framework uses a single ontology, multiple view approach that exploits modifiers and conversion functions for context mediation between different data sources [ 73 ]. To support clinical knowledge sharing through interaction models, the OpenKnowledge project has been initiated from different data sources [ 9 ], and K-MORPH architecture has been proposed for a unified prostate cancer clinical pathway.

Along with information sharing, medical data management is critical in the diagnosis of disorders like dementia. To establish a better diagnosis method for dementia, a medical information management system (MIMS) was designed using SW technologies through the extraction of metadata from medical databases [ 74 , 75 ]. In order to further eliminate the e-health information and knowledge sharing crisis, Bai and Zhang [ 76 ] suggested Integrated Mobile Information System (IMIS) for healthcare. It provides a platform to connect diabetic patients with care providers to receive proper treatment and diagnosis facilities at home. The Diabetes Healthcare Knowledge Management project also aims to ease decision support and clinical data management in diabetes healthcare processes [ 72 ].

To construct decision models for the Diabetes Healthcare Knowledge Management framework, tools such as Semantic Web Rule Language (SWRL), OWL, and RDF were used. This ontology-based knowledge framework provides ontologies, patient registries, and an evidence-based healthcare resource repository for chronic care services [ 72 ]. Web Ontology Language (OWL), Resource Description Framework (RDF), and SPARQL were also commonly used for the creation of metadata in dementia diagnosis [ 77 ]. On the other hand, the Semantic Web-based retrieval system for the pathology project, known as “A Semantic Web for Pathology,” involves building and managing ontology for lungs which was made up of common semantic tools RDF and OWL which were used along with RDQL query language [ 20 ].

Even though effective frameworks were proposed to diagnose certain diseases, research gaps still exist that affect medical data management. For instance, the fuzzy technique-based service-oriented architecture has proved to be beneficial in terms of adjustability and reliability. But still, in the context of domain-specific ontologies, the applicability of this architecture is yet to be validated [ 78 ]. Effective distribution of knowledge into the existing healthcare system is a huge challenge in augmenting decision making and improving the care service quality. Therefore, future works are intended to focus on embedding knowledge and conducting user evaluations for better disease management.

3.3. Information Management

Managing patients' information and storing test results are significant tasks in the medical and healthcare industries. The application of the SW-based approach in this area can make an influential impact on this data organization. Such an approach to gather valuable and new medical information was primarily made by creating a network of computers [ 79 ]. Domain ontology was created according to the user's choice, suggesting medical terminologies to retrieve customized medical information [ 80 ]. RDF datasets can be used to find the trustworthiness of intensive care unit (ICU) medical data [ 70 ]. The SW has also been used to document healthcare video contents [ 81 ] and radiological images to provide appropriate information about those records [ 82 ].

However, moving from the conventional web-based information management to the Semantic Web had some reasons. As medical knowledge is essential to verify and share across hospitals and medical centers, introducing the Semantic Web approach helped to achieve a proper mapping system [ 83 ]. A medical discussion forum based on the SW helped to exchange valuable data among healthcare practitioners to map-related information in the dataset [ 84 ]. The use of the fuzzy cognitive system in the SW also helped to share and reuse knowledge from databases and simplify maintenance [ 85 ]. This methodology also helped to improve data integration, analysis, and sharing between clinical and information systems and researchers [ 86 ]. Moving towards this approach also aided the researchers in connecting different data storage domains and creating effective mapping graphs [ 87 ].

Though the approach of SW in healthcare has a broad area, most applications are pretty similar. The framework mainly proposed the use of RDF, SPARQL, and OWL [ 4 , 76 ]. Link relevance methods were used to produce semantically relevant results to extract pertinent information from domain knowledge [ 49 ]. Ontology-based logical framework procedures and SMS architecture helped to organize the heterogeneous domain network [ 88 , 89 ].

Evaluating the system's performance is necessary to get the actual results. A Health Level 7 (HL7) messaging mechanism has been developed for mapping the generated Web Service Modeling Ontology [ 90 ]. However, there were some issues regarding the heterogeneity problem. JavaSIG API was used to generate the HL7 message to resolve these issues [ 91 ]. Some of the evaluation tools are not advanced enough to handle vast amounts of data. PMCEPT physics algorithms were used to verify the algorithm [ 92 ]. Abidi and Hussain [ 9 ] created two levels to characterize different ontological models to establish morphing. BioMedLib Search Engine creation for economic efficiency helped to develop a Semantic Web framework for rural people [ 25 ]. The Metamorphosis installation wizard converted the text format UMLS into a MySQL database UMLS in order to access a SPARQL endpoint [ 93 ].

However, the frameworks proposed in different statements were not implemented precisely, which created a gap in each framework. Some frameworks are proposed to integrate with the blockchain for additional security and privacy [ 23 , 94 – 96 ]. AI and IoT integration can also enhance system maintenance [ 1 ]. Hussain et al. [ 97 ] suggested a framework named Electronic Health Record for Clinical Research (EHR4CR), but they did not get any actual results from this framework in the real world [ 97 ]. The proposed framework's implementation result will provide more development on this.

3.4. Frontier Technology

In this segment, we critically analyze works that are primarily keen on how cutting-edge technologies like AI and computer vision can be applied to the medical field with the continuous advancement of science and technology. Semantic Web-enabled intelligent systems leverage a knowledge base and a reasoning engine to solve problems, and they can help healthcare professionals with diagnosis and therapy. They can assist with medical training in a resource-constrained environment. To illustrate, Haque et al. [ 8 ], Chondrogiannis et al. [ 98 ], Haque and Bhushan [ 99 ], and Haque et al. [ 24 ] created a secure, fast, and decentralized application that uses blockchain technologies to allow users and health insurance organizations to reach an agreement during the implementation of the healthcare insurance policies in each contract. To preserve the formal expression of both insured users' data and contract terms, health standards and Semantic Web technologies were used. Accordingly, significant work has been proposed by Tamilarasi and Shanmugam [ 100 ] which explores the relationship between the Semantic Web, machine learning, deep learning, and computer vision in the context of medical informatics and introduces a few areas of applications of machine learning and deep learning algorithms. This study also presents a hypothesis on how image as ontology can be used in medical informatics and how ontology-based deep learning models can help in the advancement of computer vision.

The real-world healthcare datasets are prone to missing, inconsistent, and noisy data due to their heterogeneous nature. Machine learning and data mining algorithms would fail to identify patterns effectively in this noisy data, resulting in low accuracy. To get these high-quality data, data preprocessing is essential. Besides, RDF datasets representing healthcare knowledge graphs are very important in data mining and integrating IoT data with machine learning applications [ 8 , 101 ]. RDF datasets are made up of a distinguishable RDF graph and zero or more named graphs, which are pairings of an IRI or blank node with an RDF graph. While RDF graphs have formal model-theoretic semantics that indicate which world configurations make an RDF graph true, there are no formal semantics for RDF datasets. Unlike traditional tabular format datasets, RDF datasets require a declarative SPARQL query language to match graph patterns to RDF triples, which makes data preprocessing more crucial. In the context of data preprocessing, Monika and Raju [ 101 ] proposed a cluster-based missing value imputation (CMVI) preprocessing strategy for preparing raw data to enhance the imputed data quality of a diabetes ontology graph. The data quality evaluation metrics R2, D2, and root mean square error (RMSE) were used to assess simulated missing values.

Nowadays, question-answering (QA) systems (e.g., chatbots and forums) are becoming increasingly popular in providing digital healthcare. In order to retrieve the required information, such systems require in-depth analysis of both user queries and records. NLP is an underlying technology, which converts unstructured text into standardized data to increase the accuracy and reliability of electronic health records. A Semantic Web application has been deployed for question-answering using NLP where users can ask questions about health-related information [ 27 ]. In addition, this study introduces a novel query simplification methodology for question-answering systems, which overcomes issues or limitations in existing NLP methodologies (e.g., implicit information and need for reasoning).

The majority of contributions to this category have organized their work using semantic languages on a smaller scale. Besides, it is noteworthy that hardly any of the approaches, except [ 27 , 101 ], adopted a framework for developing their models. Asma Ben et al. used a benchmark (corpus for evidence-based medicine summarization) to evaluate the question-answering (QA) system and analyzed the obtained outcomes [ 27 ]. Some studies have not included a prior literature review for the discovery of available frontier services [ 100 ]. In addition, the study shows that with the soaring demand for better, speedier, more accurate, and personalized patient treatment, deep learning powered models in production are becoming increasingly prevalent. Often these models are not easily explainable and prone to biases. Explainable AI (XAI) has grown in popularity in healthcare due to its extraordinary success in explaining decision-making criteria to systems, reducing unintended outcomes and bias, and assisting in gaining patients' trust—even when making life-or-death decisions [ 102 ]. To the best of our knowledge, XIA has gleaned attention on ontology-based data management but received relatively little attention on collaborating Semantic Web technologies across healthcare, biomedical, clinical research, and genomic medicine. Similarly, within the IoT system spectrum, invocation of semantic knowledge and logic across various Medical Internet of Things (MIoT) applications, gathering vast amounts of data, monitoring vital body parameters, and gathering detailed information from sensors and other connected devices, as well as maintaining safety, data confidentiality, and service availability also received relatively little attention.

3.5. Regulatory Conditions

This segment concentrates on Semantic Web-based tools, technologies, and terminologies for documenting the semantics of medical and healthcare data and resources. As the healthcare industries generate a massive amount of heterogeneous data on a global scale, the use of a knowledge-based ontology on such data can reduce mortality rate and healthcare costs and also facilitate early detection of contagious diseases. Besides, the SW provides a single platform for sharing and reusing data across apps, companies, and communities. The biomedical community has specific requirements for the Semantic Web of the future. There are a variety of languages that can be used to formalize ontologies for medical healthcare, each with its expressiveness. A collaborative effort led by W3C, involving many research and industrial partners, set the requirements of medical ontologies. A real ontology of brain cortex anatomy has been used to assess the requirements stated by W3C in two available languages at that time, Protégé and DAML + OIL [ 103 ]. The development and comparative analysis contexts of brain cortex anatomy ontologies are partially addressed in this. In 2019, a survey-based study was conducted to determine faculty and researcher usage, impact, and satisfaction with Web 3.0 networking sites on medical academic performance [ 104 ]. This study explores the awareness and willingness to implement Web 3.0 technologies within healthcare at Rajiv Gandhi University of Health Sciences. The results of this study imply that Web 3.0 technologies have an impact on professor and researcher academic performance, with those who are tech-savvy being disproportionately found in high-income groups [ 104 ].

Documentation of semantic tools and data is required to resolve healthcare reimbursement challenges. Besides, regulations are also necessary to standardize semantic tools while ensuring that healthcare communities and systems adhere to general health policies. Unfortunately, we found only a few works focusing on this challenge based on SWT. Only the study conducted by Sugihartati [ 104 ] adopted a proper survey methodology. Therefore, future efforts should focus on regulating, documenting, and standardizing semantic tools, technologies, and health resources, as well as conducting user evaluations to understand and optimize functional efficiency and accelerate market access for medicines for general health.

Tables ​ Tables3 3 ​ 3 – 5 provide a detailed analysis of the studied works for the derived five categories.

Summarization of the research contribution of the selected articles.

Summarization of the research gaps and future research avenues.

Future research avenues in the form of research questions.

4. Research Gaps

This systematic literature review presents a vast knowledge about the use of Web 3.0 or Semantic Web technology in different approaches to the medical and healthcare sector. By analyzing various kinds of literature, we recognized different research gaps to address future research avenues, which will enable scholars from different parts to examine the area and discover new developments. Table 4 summarizes the overall research gaps and Table 5 summarizes the future research avenues we encounter during the literature review.

4.1. Scope of E-Healthcare Service Research

Even though studies in the domain of e-healthcare services suggested and created numerous frameworks to provide vital support to the users, there are still research gaps among the methods. Several frameworks were proposed to facilitate data interoperability. However, based on what we know best, none of the proposed frameworks has been implemented in the actual world. Furthermore, there is no evidence of knowledge sharing among organizations using semantic network-based systems. Besides, just a handful of the research papers included assessment methodologies and a discussion of the findings. Furthermore, the frameworks that provide medical services such as disease reasoning, decision making, and drug recommendations lack reliability and validity. Most of the research articles suggested architectures but did not implement them, and their intended prototypes were never built.

4.2. Scope of Disease Research

Semantic Web technologies are being used in the healthcare sector to improve information interoperability and aid in identifying and treating diseases. Only a few studies among the 65 papers have examined the various frameworks for developing a fully functional system for either diabetic healthcare or disease collection of prebuilt queries. Earlier research also lacks mapping triplets of one illness RDF to other existing medical services, applications, and administrations. Researchers also lack the creation of intelligent user interfaces that grasp the semantics of clinical data. This paper shows that more study is required to efficiently use ontology in the healthcare sector to preserve data with proper evaluation criteria.

4.3. Scope of Information Management Research

Medical data are considered valuable information utilized to assist patients in receiving better care. It is challenging to implement Semantic Web technologies to store and search for data. Various studies attempt to adopt specific methods that may aid in the proper management of medical information; however, some gaps remain. There is no attempt to index high-quality videos and collect attributes for categorizing them. A validation gap exists due to the lack of suitable evaluation techniques. In most studies, RDF ontologies are used to collect information from websites and represent those data. However, no information is provided about how effective those models are in real-world applications.

4.4. Scope of Frontier Technology Research

Even though cutting-edge technology such as AI, ML, robotics, and the IoT has revolutionized the healthcare industry and helped improve everything from routine tasks to data management and pharmaceutical development, the industry is still evolving and looking for ways to improve. If we consider the aspect of research, the history of the Semantic Web and frontier technology is technically not new at all, yet the Semantic Web presents some limitations. Since the web began as a web of documents, converting each document into data is incredibly challenging. Various tools and approaches, such as natural language processing (NLP), may be used to do this task, but it would take a long time. However, only a small attempt has been made to integrate NLP and domain knowledge induction. Ontology and AI, and logic, have always been and will continue to be essential elements of AI development. Besides, connecting ontology to AI is frequently a problem in and of itself. Furthermore, because ontology trees often have a large number of nodes, real-time execution is problematic. Earlier studies have apparently failed to solve this problem. There have been significant attempts to incorporate the various aspects of IoT resources into ontology creation, such as connectivity, virtualization, mobility, energy, or life cycle [ 108 , 109 ]. The authors attempted to enhance the computerization of the health and medical industry by utilizing the Internet of Things (IoT) and Semantic Web technologies (SWTs), which are two key emerging technologies that play a significant role in overcoming the challenges of handling and presenting data searches in hospitals, clinics, and other medical establishments on a regular basis. Despite its significant efforts to collaborate different IoT spectrum and Semantic Web technologies, research gaps in medical data management persist. For instance, after its introduction, the Medical Internet of Things (MIoT) has taken an active role in improving the health, safety, and care of billions of people. Rather than going to the hospital for help and support, patients' health-related parameters can now be monitored remotely, constantly, consistently, and in real time and then processed and transferred to medical data enters via cloud storage. Because of cloud platforms' security risks, choosing one is a major technological challenge for the healthcare industry. Some of these cloud-based storage systems cannot adequately preserve patients' data and information regarding semantic data [ 6 , 8 ]. However, none of the research articles suggested any architectures, nor were any intended prototypes built to address these cloud security issues of MIoT in general.

4.5. Scope of Regulatory Condition Research

Regulations are paramount for the healthcare and medical industries to function properly. They support the global healthcare market, ensure the delivery of healthcare services, and safeguard patients,' doctors,' developers,' researchers,' and healthcare agents' rights and safety. The Semantic Web also has its detractors, like many other technologies, in terms of legislation and regulation. Historically, scaling medical knowledge graphs has always been a challenge. As a result of privacy and legal clarity, healthcare companies are not sufficiently incentivized to share their data as linked data. Only a few academic papers and documents disclose how these corporations use to automate the process. Furthermore, compared to other types of datasets, many linked datasets representing tools are of poor quality. As a result, applying them to real-world problems is highly challenging. Other alternatives, such as property graph databases like Neo4j and mixed models like OriendDB, have grown in popularity due to the RDF format's complexity. Healthcare application developers and designers prefer to use web APIs over SPARQL endpoint to send data in JSON format. This study illustrates that more research is needed to improve the semantic quality of available technologies (e.g., RDF, OWL, and SPARQL) to effectively use them in the healthcare industry to ease healthcare development.

5. Discussion

This section describes the findings from the selected studies based on answer to the research questions. Therefore, the readers will be able to map the research questions with the contribution of this systematic review.

5.1. (RQ1) What Is the Research Profile of Existing Literature on the Semantic Web in the Healthcare Context?

The research aims to determine the primary objectives of using the Semantic Web and the major medical and healthcare sectors where Semantic Web technologies are adopted. As the Semantic Web has shown incremental research trends in recent years, there is a need for a structured bibliometric study. This study collected data from the Scopus, IEEE Xplore Digital Library, ACM Digital Library, and Semantic Scholar databases, focusing on various aspects and seeing their affinity. We performed bibliometric analysis to look at essential details like preliminary information, country, author, and application area where these publications are being used for the Semantic Web in the context of healthcare. We conducted the bibliometric analysis using an open-source application called VOS viewer. The outcomes and specifics of the experiment are detailed in Section 2 .

As stated in the methodology section, our study consists of 65 documents. A number of prestigious conferences, publications, and events have published these healthcare-related articles. Out of these 65 shortlisted papers, 27 were presented in conferences, 21 in journals, and 17 from book chapters. Our study observes that the field of “Semantic Web in Healthcare” is not comparatively new. The first paper from the shortlisted documents on this topic was published in 2001. Since then, there has been minimal growth in this field, with 2007 appearing to be the start. Surprisingly, the maximum number of articles (8) published in this discipline was in 2013, but from 2013 to 2016, there was only a minor shift by researchers globally. It is most likely due to the introduction of Web 3.0 in 2014. It is yet to be found how Web 3.0 will effectively leverage the Semantic Web as a core component rather than seeing it as a competing technology in the medical healthcare field. The decrease in the number of articles shows how the interests of researchers switched from the Semantic Web to the emerging Web 3.0. However, the Semantic Web remains the top choice of medical practitioners as Web 3.0 evolves. Furthermore, the United States is the country with the most research papers, followed by France and India (see Figure 4 ). It implies that both developed and emerging countries use the Semantic Web in their healthcare industries. VOS viewer also discovered 35 works titled to be published in Computer Science, 16 in Engineering, 9 in Medicine, and 5 in Mathematics. We also used the VOS viewer software to visually represent the keyword co-occurrences from those shortlisted 65 publications. The total number of keywords was 774. The minimum number of times a keyword appears is set at 5. The terms that occurred more than five times in all texts are included in our representation. We found 76 keywords that meet our requirements. Figure 7 shows our findings in a co-occurrence graph containing the other essential phrases. As expected, Semantic Web and healthcare are the most occurring keywords, and both are mentioned 55 times. Following that, web services, decision support systems, interoperability, etc. are listed. These terms are used to categorize the Semantic Web's application areas in healthcare.

Co-occurrence network of the index's keywords.

Our analysis also reveals that most proposed frameworks for improving and expanding the healthcare system do so without the involvement of health professionals. Some of them discussed data interoperability, diseases, frontier technologies, and regulatory issues, while others emphasized the use of video as ontologies and video conferences in bridging communication gaps. The majority of the publications only propose frameworks with no implementation. Web services currently merely make services available, with no automatic mechanism to connect them in a meaningful way.

5.2. (RQ2) What Are the Primary Objectives of Using the Semantic Web, and What Are the Major Areas of Medical and Healthcare Where Semantic Web Technologies Are Adopted?

The adoption of the Semantic Web in healthcare strives to improve collaboration, research, development, and organizational innovation. The Semantic Web has two primary objectives: (1) facilitating semantic interoperability and (2) providing end-users with more intelligent support. Semantic interoperability, a key bottleneck in many healthcare applications, is one of today's major problems. Semantic Web technologies can help with data integration, knowledge administration, exchange of information, and semantic interoperability between healthcare information systems. It focuses on building a web of data and making it appropriate for machine processing with little to no human participation. So, healthcare computer programs can better assist in finding information, personalizing healthcare information, selecting information sources, collaborating within and across organizational boundaries, and so on by inferring the consequences of data on the Internet.

Based on our review of the findings, we found five application domains where the Semantic Web is being adopted in the healthcare context. This study will brief those domains from Sections 5.2.1 to 5.2.5 as well as justify them in relation to healthcare.

5.2.1. E-Healthcare Service

More than two-fifth of the total studies (65) considered in this study is about e-healthcare services (see Table 2 ). These studies focus on ways to use the Internet and related technologies to offer and promote health services and information, as well as diagnosis recommendation systems and online healthcare service automation.

In this study, researchers developed a web-based prototype that generates the required reports with a high degree of data integration and a rule-based production technique for establishing a link between prevalent diseases and the range of diseases in a specific gene [ 64 ].

Another group of e-healthcare service studies focused on how current electronic information and communication technology could help people's health and healthcare [ 46 , 49 , 50 , 61 – 64 , 97 ]. Most of the authors used a WSMO (Web Service Modeling Ontology) service delivery platform and an automatic alignment of user-defined EHR (electronic health record) workflows, where service owners can register a service, and the system will automate prefiltering, discovery, composition, ranking, and invocation of that service to provide healthcare.

The adoption of e-healthcare in developing countries has shown to be a feasible and effective option for improving healthcare. It allows easy access to health records and information and reduces paperwork, duplicate charges, and other healthcare costs. If the proper implementation of e-healthcare technologies is ensured, everyone will benefit.

5.2.2. Diseases

Out of 65 articles, there are only 8 articles regarding the adoption of the Semantic Web in the diseases sector (see Table 2 ). These articles present a discussion on the deployment of a disease-specific healthcare platform, disease information exchange system, knowledge base generation, and research portal for a specialized disease.

This study developed a web-based prototype for an Integrated Mobile Information System (IMIS) for diabetic patient care [ 20 ]. The authors used ontology mapping so that related organizations could access each other's information. They also embedded feedback and communication mechanisms within the system to include user feedback.

Another study developed queryMed packages for pharmaco-epidemiologists to access and link medical and pharmacological knowledge to electronic health records [ 10 ]. The authors distinguished all the medications endorsed for critical limb ischemia (CLI) and recognized one contraindicated solution for one patient.

Disease management/prediction systems are necessary for finding the hidden knowledge within a group of disease data and can be used to analyze and predict the future behavior of diseases. An all-in-one strategy rarely works in the healthcare industry. It is critical to develop a personalized and contextualized disease prediction system to enhance user experience.

5.2.3. Information Management

Almost two-fifths of the total studies considered in this study (65) are about information management (see Table 2 ). After e-healthcare service, this category has the most studies. These articles are particularly about healthcare management systems, medical information indexing, healthcare interoperability systems, decision making, coordination, control, analysis, and visualization of healthcare information.

This study presented a medical knowledge morphing system that focuses on ontology-based knowledge articulation and morphing of heterogeneous information using logic and ontology mediation [ 105 ]. The authors used high-level domain ontology to describe fundamental medical concepts and low-level artifact ontology to capture the content and structure.

In another study, an annotation image (AIM) ontology was developed to provide important semantic information within photographs, allowing radiological images to be mined for image patterns that predict the structures' biological features. The authors transformed XML data into OWL and DICOM-SR to control ontological terminology in order to create image annotation.

A well-designed healthcare information system is required for management, evaluation, observations, and overall quality assurance and improvement of key stakeholders of the health system. Even though a significant amount of work is done in this sector, it is far from sufficient. It is something on which we should focus.

5.2.4. Frontier Technology

We found only 3 publications on frontier technology (see Table 2 ). These articles describe healthcare application domains that use AI, machine learning, or computer vision to automate medical coding, generate medical informatics, and deal with intelligent IoT data and services.

The first review article is about a method for preprocessing raw cluster-based missing value imputation (CMVI), with the goal of improving the imputed data quality of a diabetes ontology graph [ 27 ]. Their findings show that preprocessed data have better imputation accuracy than raw, unprocessed data, as measured against coefficient of determination (R2), index of agreement (D2), and root mean square error (RMSE).

Another article talks about ideas on how image as ontology can be used in health informatics and how deep learning models built on ontologies can support computer vision [ 100 ].

Frontier technology such as AI, ML, and IoT offers many advantages over traditional analytics and clinical decision-making methodologies. At a granular level, those technologies provide

• Increased efficiency.
• Better treatment alternatives.
• Faster diagnosis.
• Faster drug discovery.
• Better disease outbreak prediction.
• Medical consultations with patients with little or no participation of healthcare providers.

There is a lack of research on the integration of frontier technologies with the Semantic Web. Researchers should focus their efforts on this area. Students must take the initiative to develop creative technological inventions.

5.2.5. Regulatory Conditions

There were only 3 publications that used Semantic Web technology to address regulatory conditions (see Table 2 ). These studies focus on the challenges and requirements of the Semantic Web and technologies that represent the Semantic Web, awareness, and policy and regulations.

An article describes how to design, operate, and extend a Semantic Web-based ontology for an information system of pathology [ 103 ]. The authors of this paper highlight what technologies, regulations, and best practices should be followed during the entire lung pathology knowledge base creation process.

Another study talks about the challenges of integrating healthcare web service composition with domain ontology to implement diverse business solutions to accomplish complex business logic [ 104 ].

Privacy and regulation are important in establishing a clear framework within which healthcare providers, patients, healthcare agents, and healthcare application developers can learn and maintain the skills needed to provide high-quality health services which are safe, productive, and patient-centered. From these regulatory condition-type articles, we can understand whether technology is easy to use, has challenges, and is emerging, secure, and valuable to the healthcare community. We need to do more work on this.

5.3. (RQ3) Which Semantic Web Technologies Are Used in the Literature, and What Are the Familiar Technologies Considered by Each Solution?

This section discusses the various Semantic Web technologies used in the literature, as well as the most common ones among them. There are numerous Semantic Web technologies available that make the applications more advanced. The healthcare industry makes extensive use of these Semantic Web technologies. As a result of these technologies, the healthcare industry is getting more advanced. The most prevalent Semantic Web technologies that are used in the healthcare sector are Resource Description Framework (RDF), Web Ontology Language (OWL), SPARQL Protocol and RDF Query Language (SPARQL), Semantic Web Rule Language (SWRL), Web Service Modeling Ontology (WSMO), Notation3 (N3), SPARQL Inferencing Notation (SPIN), Euler Yap Engine (EYE), Web Service Modeling Language (WSML), and RDF Data Query Language (RDQL).

Various Semantic Web technologies are used to accomplish various goals, such as converting relational databases to RDF/OWL-based databases, data linking, reasoning, data sharing, data representation, and so on. Ontologies are considered the basis of the Semantic Web. All of the data on the Semantic Web are based on ontologies. To take advantage of ontology-based data, it must first be transformed into RDF-based datasets. The RDF is an Internet standard model for data transfer that includes qualities that make data merging easier, as well as the ability to evolve schemas over time without having to update all of the data [ 52 ]. The majority of the researchers utilized RDF to represent the linked data and interchange data. In the Semantic Web, Notation3 is used as an alternative to RDF to construct notations. It was created to serialize RDF models and it supports RDF-based principles and constraints. Humans can understand Notation3-based notations more easily than RDF-based notations. In addition to RDF, OWL is employed in the research articles to express ontology-based data. The OWL is a semantic markup language for exchanging and distributing ontologies on the web [ 52 ]. Furthermore, there is a second version of OWL available which is known as OWL2. The improved descriptive ability for attributes, enhanced compatibility for object types, simplified metamodeling abilities, and enhanced annotation functionality are among the new features added in OWL2. Numerous OWL-based ontologies are available on the web. OWL-S is one of them which is a Semantic Web ontology [ 78 ]. The OWL is also used for semantic reasoning. Combining Description Logic with OWL (OWL-DL) takes the reasoning capability to another level. OWL-DL provides desired algorithmic features for reasoning engines and is meant to assist the current Description Logic industry area [ 82 ]. As an alternative to OWL, EYE is used which is an advanced chaining reasoner with Euler path detection [ 85 ]. It uses backward and forward reasoning to arrive at more accurate conclusions and results. To query the RDF and OWL-based datasets, the scholars made use of SPARQL. SPARQL is the sole query language that may be used to query RDF and OWL-based databases. However, RDQL was employed as a query language for RDF datasets in a study [ 20 ]. Only RDF datasets can be queried with it. In several papers, writing the semantic rules and constraints was necessary. So, they used SWRL which is a language for writing semantic rules based on OWL principles. Alongside SWRL, scholars used SPIN which is a rule language for the Semantic Web that is based on SPARQL [ 60 ]. In the Semantic Web, specifying web services for different purposes is essential. In this regard, some research papers discussed leveraging the WSMO which is a Semantic Web framework for characterizing and specifying web services in a semantic way. A linguistic framework called WSML is used to express the Semantic Web services specified in WSMO. The WSML is a syntactic and semantic language framework for describing the elements in WSMO [ 48 ]. Tables ​ Tables4 4 ​ 4 ​ ​ – 8 summarize the Semantic Web technologies employed in different thematic research areas. Section 3 has detailed information regarding the discussion.

Summary of Semantic Web technologies used in e-healthcare services.

Summary of Semantic Web technologies used in diseases.

Summary of Semantic Web technologies used in information management.

Table 6 summarizes Semantic Web technologies used in e-healthcare services. In this field of theme research, RDF, OWL, and SPARQL are the most commonly utilized technologies. Researchers employed RDF and OWL to construct RDF-based datasets, represent RDF datasets, and develop links between data. As an alternative to RDF, an article used Notation3 to construct RDF notations which are easier to read than RDF-based notations. In a paper, the scholars used OWL2, the second version of OWL, to utilize the latest features offered by the technology. For all of the articles, SPARQL is the only query language utilized to query the datasets. To construct rules and limits for the systems, most of the articles used SWRL. In addition to SWRL, an article used the SPIN to generate semantic rules and constraints. Furthermore, SPIN has not been used in any other research area. Besides, two articles used WSMO for the identification of Semantic Web services required for the systems. On the other hand, three articles in this theme did not use any Semantic Web technology.

Table 7 summarizes Semantic Web technologies used in diseases. Similar to the preceding thematic research area, RDF, OWL, and SPARQL are the most frequently used technologies. Also, the motivations for using these technologies are identical. However, an article utilized RDQL as an alternative to SPARQL to conduct queries on RDF datasets. SWRL was used to construct rules and limitations, just as it had been previously. It is also worth noting that a study built a model using the OWL-S, an OWL-based semantic ontology. Then, there is a study in this field that did not utilize any Semantic Web technology at all.

Table 8 summarizes Semantic Web technologies used in information management. Nine distinct Semantic Web technologies are used in this thematic research area. RDF, OWL, and SPARQL, like the previous topic groups, are the most extensively used technologies. It is worth repeating that the technologies' goals are the same as they were previously. In addition, the usage of Notation3 for more accessible RDF notations, OWL2 to take advantage of new capabilities, OWL-S semantic ontology as the data source, and WSMO to identify Semantic Web services are also mentioned in this thematic area. In this field of research, there are two new technologies that are not present in prior fields. OWL-DL, which combines OWL with Description Logic for information reasoning, is one of the new technologies. The other one is EYE reasoner, which is also a reasoning engine. On the contrary, a significant proportion of articles, six to be exact, did not employ any Semantic Web technologies.

Table 7 summarizes Semantic Web technologies used in frontier technology. In this thematic study field, there are just three articles, and two of them did not employ any kind of Semantic Web technology. The other paper includes RDF and SPARQL, which were very commonly used in the prior thematic research fields.

Table 8 summarizes Semantic Web technologies used in regulatory conditions. Only one of the two articles in this research area includes Semantic Web technology. Also, the sole semantic technology used in the article is RDF for the purpose of semantic data representation.

There are different applications of Semantic Web technologies in the articles, but most of the technologies are common in several articles. The most commonly used Semantic Web technologies are the SPARQL query language, RDF, OWL, and SWRL. Almost 80 percent of the analyzed papers used different functionalities of RDF. Furthermore, OWL and SPARQL technologies were used in nearly three-quarters of the articles. Besides, SWRL technology was applied in one-third of the analyzed studies. It is now obvious that these technologies have the potential to improve the healthcare industry.

5.4. (RQ4) What Are the Evaluating Procedures Used to Assess the Efficiency of Each Solution?

The suggested technologies and procedures for evaluating these works are included in this category. In truth, assessing the designed healthcare system's quality, performance, and utility is a crucial responsibility. Because the healthcare industry is highly sensitive, suitable evaluation standards are necessary. Due to technological limitations, however, the evaluation system is not well organized or maintained. Because the notion of Semantic Web technology is new in the medical field, overall development and evaluation are inadequate.

In the e-healthcare service-based theme (see Table 6 ), the authors in [ 51 ] established a set of setups to test the matcher's efficiency for scalability in terms of the number of Semantic Web services for medical appointments and their complexity. They consider the logical complexity of Flora-2 expressions used in pre and post-conditions, which can handle various web service and goal descriptions, including ontology consistency check. Some other evaluation methods like OSHCO validation for automatic decision support in medical services were also introduced by the authors in [ 57 ]. An experiment was established to assess the system utilizing two metrics via WS datasets, the execution time measurement and the correctness measurement, for graph-based Semantic Web services for healthcare data integration [ 62 ] and histopathology for evaluating the performance of semantic mappings [ 58 ].

However, only two publications presented evaluation procedures from the vast portion of information management system-related work (see Table 7 ). Tonguo et al. [ 25 ] used BioMedLib to evaluate a system that takes a user's search query and pulls articles from millions of national biomedical article databases. Another one used evaluation criteria like D2RQ for default semantic mapping generation [ 83 ].

In terms of frontier technology (see Table 9 ), the cluster-based missing value imputation algorithm (CMVI) was used to extract knowledge in the Semantic Web's healthcare domain [ 101 ]. The imputation accuracy was measured using a couple of well-known performance metrics, namely, coefficient of determination (R2) and index of agreement (DK), along with the root mean square error (RMSE) test. In addition, various open-domain question-answer evaluation campaigns such as TREC21, CLEF22, NTCIR23, and Quaero24 have been launched to evaluate a Semantic Web and NLP-based medical questionnaire system [ 27 ].

Summary of Semantic Web technologies used in frontier technology.

None of the writers provide any evaluation methodologies connected to diseases and regulatory conditions (see Tables ​ Tables9 9 and ​ and10). 10 ). To assess the consequences of Semantic Web discussions on specific diseases, well-designed evaluation criteria are required. As studies focus on the obstacles and problems of the Semantic Web in healthcare services, the necessity of evaluation is also missing in regulatory conditions.

Summary of Semantic Web technologies used in regulatory conditions.

5.5. (RQ5) What Are the Research Gaps and Limitations of the Prior Literature, and What Future Research Avenues Can Be Derived to Advance Web 3.0 or Semantic Web Technology in Medical and Healthcare?

The healthcare industry is on the verge of a real Internet revolution. It intends to bring in a new era of web interaction through the adoption of the Semantic Web, with significant changes in how developers and content creators use it. This web will make healthcare web services, applications, and healthcare agents more intelligent and even provide care with human-like intelligence by utilizing an AI system. Despite the tremendous amount of innovation, it may bring its adoption in healthcare considerable challenges.

The problem with the “Semantic Web” is that it requires a certain level of implementation commitment from web developers and content creators that will not be forthcoming. First, a large portion of existing healthcare web content does not use semantic markup and will never do so due to a lack of resources to rewrite the HTML code. Second, there is no guarantee that new healthcare content will utilize semantic markup because it would need additional effort. However, it is essential to guide the Semantic Web developer community in the right direction so that they can help contribute to future medical healthcare development. The following are the primary obstacles the Semantic Web faces in general: (i) content availability, (ii) expanding ontologies, (iii) scalability, (iv) multilingualism, (v) visualization to decrease information overload, and (vi) Semantic Web language stability.

Furthermore, based on our thorough examination of the 65 publications, the following are some of the most technologically severe obstacles that the Semantic Web in general faces in the healthcare context and must overcome; future research may be able to alleviate a few of these challenges:

• Integrated Data Issue . The vulnerability of interconnected data is one of the most significant challenges with Semantic Web adoption. All of a patient's health records and personal information are stored and interlinked to an endpoint, and a malicious party may gain control of one's life if the record is compromised.
• Vastness . The current Internet contains a vast amount of healthcare records not yet semantically indexed; any reasoning system that wants to analyze all of these data and figure out how it functions will have to handle massive amounts of data.
• Vagueness . As Semantic Web is not yet mature enough, applications cannot handle non-specific user queries adequately.
• Accessibility . Semantic Web may not work on older or low-end devices; only highly configured devices will be able to manage web content.
• Usability . It will be difficult for beginners to comprehend because the SPARQL queries are often used in websites and services.
• Deceit . What if the information provided by the source is false and deceptive? Management and regulation have become crucial.

The study also identifies future research opportunities and gives research recommendations to the developer and researcher communities for each of the identified theme areas where the Semantic Web is being used in medical and healthcare (see Section 4 ). Tables ​ Tables4 4 and ​ and5 5 summarize the research gap and probable future research direction.

6. Conclusion

The purpose of this SLR is to discover the most recent advances in SW technology in the medical and healthcare fields. We used well-established research techniques to find relevant studies in prestigious databases such as Scopus, IEEE Xplore Digital Library, ACM Digital Library, and Semantic Scholar. Consequently, we were able to answer five significant RQs. We answered RQ1 by giving a bibliometric analysis-based research profile of the existing literature. The study profile includes information on annual trends, publishing sources, methodological approaches, geographic coverage, and theories applied (see Sections 2.4 and 5.1 ). We performed content analysis to determine the answers to RQ2 , RQ3 , and RQ4 ; we also identified research themes, with a focus on technical challenges in healthcare where SW technologies can be used (see Sections 3 and 5.2 – 5.4 ). Finally, the synthesis of prior literature helped us to identify research gaps in the existing literature and suggest areas for future research in RQ5 (see Section 5.5 and Tables ​ Tables4 4 and ​ and5). 5 ). The findings of this study have important implications for healthcare practitioners and scholars who are interested in the Semantic Web and how it might be used in medical and healthcare contexts.

The global digital healthcare market is growing to meet the health needs of society, individuals, and the environment. As a result, a substantial study is required to assist governments and organizations in overcoming technological challenges. We successfully reviewed 65 academic papers comprising journal articles, conference papers, and book chapters from prestigious databases. We have identified five thematic areas based on our research questions to discuss the objectives, solutions, and prior work of Semantic Web technology in the healthcare field. Among these, we observed that e-healthcare services and medical information management are the most discussed topics [ 105 , 107 ]. According to our findings, with the emergence of Semantic Web technology, integration, discovery, and exploration of medical data from disparate sources have become more accessible. Accordingly, medical applications are incorporating semantic technology to establish a unified healthcare system to facilitate the retrieval of information and link data from multiple sources. Most of the studies that we examined discussed the importance of knowledge sharing among clinicians and patients to develop an effective medical service. The frameworks described depended on the proper data distribution from various sources supported by specific technology interventions [ 24 ]. To answer patient queries, SW-based systems such as appointment matchmaking, quality assurance, and NLP-based chatbots have been proposed to improve healthcare services [ 24 , 111 , 112 ]. In short, the Semantic Web has huge potential and is widely regarded as the web's future, Web 3.0, which will present a new challenge and opportunity in combining healthcare big data with the web to make it more intelligent [ 6 , 113 ].

The analysis of the proposed solutions discussed in the papers helped us to identify the main challenges in healthcare systems. Besides that, this study also identifies future challenges and research opportunities for future medical researchers. We observed that most of the proposed solutions are yet to be implemented and many problems are only rudimentarily tackled so far. In conclusion, by exchanging knowledge among physicians, researchers, and healthcare professionals, the SW encourages improvement from the “syntactic” to “semantic” and finally to the “pragmatic” level of services, applications, and people. From the overall observation of the findings of this SLR, a future strategy will be to adopt some of the suggested solutions to overcome the shortcomings and open a new door for the medical industry. In the future, we will try to implement such solutions and eliminate the problems.

Data Availability

Conflicts of interest.

The authors declare that they have no conflicts of interest.

A new look of the Semantic Web

Research areas.

Data Management

Data Mining and Modeling

Learn more about how we conduct our research

We maintain a portfolio of research projects, providing individuals and teams the freedom to emphasize specific types of work.

Academia.edu no longer supports Internet Explorer.

To browse Academia.edu and the wider internet faster and more securely, please take a few seconds to  upgrade your browser .

•  We're Hiring!
•  Help Center

Semantic Web Services

• Most Cited Papers
• Most Downloaded Papers
• Newest Papers
• Last »
• Semantic Web Follow Following
• Semantic Web Technologies Follow Following
• Web Services Follow Following
• Linked Data Follow Following
• Semantic Web technology - Ontologies Follow Following
• Web Services Composition Follow Following
• Web Science Follow Following
• Data Mining in Bioinformatics Follow Following
• Computational Creativity Follow Following
• Ontology (Computer Science) Follow Following

Enter the email address you signed up with and we'll email you a reset link.

• Academia.edu Journals
•   We're Hiring!
•   Help Center
• Find new research papers in:
• Health Sciences
• Earth Sciences
• Cognitive Science
• Mathematics
• Computer Science
• Academia ©2024

IEEE Account

• Change Username/Password
• Update Address

Purchase Details

• Payment Options
• Order History
• View Purchased Documents

Profile Information

• Communications Preferences
• Profession and Education
• Technical Interests
• US & Canada: +1 800 678 4333
• Worldwide: +1 732 981 0060
• Contact & Support
• About IEEE Xplore
• Accessibility
• Terms of Use
• Nondiscrimination Policy
• Privacy & Opting Out of Cookies

A not-for-profit organization, IEEE is the world's largest technical professional organization dedicated to advancing technology for the benefit of humanity. © Copyright 2024 IEEE - All rights reserved. Use of this web site signifies your agreement to the terms and conditions.

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

• View all journals
• Explore content
• About the journal
• Publish with us
• Sign up for alerts
• Data Descriptor
• Open access
• Published: 07 September 2024

A Semantic Knowledge Graph of European Mountain Value Chains

• Valentina Bartalesi 1 ,
• Gianpaolo Coro   ORCID: orcid.org/0000-0001-7232-191X 1 ,
• Emanuele Lenzi 1 ,
• Nicolò Pratelli 1 ,
• Pasquale Pagano 1 ,
• Michele Moretti 2 &
• Gianluca Brunori 2  

Scientific Data volume  11 , Article number:  978 ( 2024 ) Cite this article

Metrics details

• Socioeconomic scenarios

The United Nations forecast a significant shift in global population distribution by 2050, with rural populations projected to decline. This decline will particularly challenge mountain areas’ cultural heritage, well-being, and economic sustainability. Understanding the economic, environmental, and societal effects of rural population decline is particularly important in Europe, where mountainous regions are vital for supplying goods. The present paper describes a geospatially explicit semantic knowledge graph containing information on 454 European mountain value chains. It is the first large-size, structured collection of information on mountain value chains. Our graph, structured through ontology-based semantic modelling, offers representations of the value chains in the form of narratives. The graph was constructed semi-automatically from unstructured data provided by mountain-area expert scholars. It is accessible through a public repository and explorable through interactive Story Maps and a semantic Web service. Through semantic queries, we demonstrate that the graph allows for exploring territorial complexities and discovering new knowledge on mountain areas’ environmental, societal, territory, and economic aspects that could help stem depopulation.

Background & Summary

The 2018 update of the World Urbanization Prospects, released by the United Nations Department of Economic and Social Affairs, projects a significant shift in the global population. Currently, 47% of the World population is rural, but this is expected to decline to 30% by 2050 1 . This transition raises concerns for traditional and cultural heritage, urban well-being, and ecological sustainability. Massive rural-to-urban migration will increase city populations, pollution, and energy consumption. The United Nations (UN) predicts that by 2050, 7 billion people will live in cities 2 , creating unsustainable conditions for food, health, and energy security. Challenges include excessive public administration burdens, cost of living mismatched with salaries, surges in pollution and greenhouse gas emissions, and increased healthcare expenditures. Additionally, cities will depend more on distant resources, leading to negative environmental impacts. In Europe, 36% of the territory is mountainous, critical for public and private-goods supply. Therefore, understanding factors that can mitigate rural and mountain area depopulation is crucial for sustainability. This aligns with the UN’s Sustainable Development Goal 11 2 , emphasising strategies to preserve rural economies and services, including economic diversification and tourism enhancements. Recent international initiatives promote capacity-building and participatory processes involving stakeholders and policymakers to create resilient and sustainable mountain areas in response to climate change 3 .

However, such endeavours demand substantial volumes of data to generate meaningful insights. Specifically, data related to environment, geography, demographics, and economics are essential for comprehending how regions and mountain-based value chains react and adapt to climate change. Such information is key to gaining new knowledge on the dynamics of mountain value chains. For example, it allows identifying the value chains sharing the same environmental characteristics (e.g. rivers, lakes, vineyards, chestnut trees) and issues (e.g. depopulation, emigration, climate change problems) or similar products across territories (e.g. cow or sheep milk cheese).

The present paper describes an extensive collection of data on 454 value chains from 23 European mountain areas belonging to 16 countries, representing as far as possible the diversity of European mountain areas (Supplementary Table  1 ). Although these data do not cover the entire spectrum of European mountain value chains, they aim to be as representative and reliable as possible by embedding information from local experts with comprehensive views. These experts overall selected the 454 value chains as those with the highest importance in the 23 areas from a socio-ecological (innovation, stage of development, size, governance system, and environmental status and protection) perspective. When they had sufficient information, the experts also extended the data beyond their monitored territories to connect and compare the value chains with similar or related territories and other value chains not initially involved in the project (e.g., the Scandinavian Mountains, the Massif Central, and the Pyrenees).

Data representation and publication principles

Our data collection is organised as a semantic knowledge graph , i.e., a structured representation of knowledge, where knowledge entities and their relations are modelled in a graph format 4 , 5 . The structure of our graph is based on an ontology. An ontology ( computational ontology, formally) is a model of the knowledge structure of a system in terms of the significant classes (e.g., event, person, location, object) and relations emerging from the observation of the system itself 6 . It is an abstract, simplified view of the system structured in a machine-readable format. Building an ontology requires identifying the relevant concepts and relations (symbolised by unary and binary predicates) of a domain of interest, and organising them in a hierarchical structure. Summarising, a semantic knowledge graph based on an ontology has nodes corresponding to the classes of the ontology and edges corresponding to the relations. A narrative is an example of a system describable as a knowledge graph modelled on an ontology.

Our collection is a semantic knowledge graph of narratives, where each narrative is a sub-graph explaining one among the 454 value chains. Each value chain narrative is a semantic network of narrative events related to each other through plot-dependent semantic relations. The overall graph is described under the Web Ontology Language (OWL) and complies with the Narrative Ontology 7 (NOnt), which provides a structure to represent the knowledge of a narrative formally. NOnt is used in several cultural heritage projects (e.g., CRAEFT 8 , Mingei 9 , and IMAGO 10 ). It reuses classes and properties (and complements) the CIDOC CRM ISO standard 11 and other standard ontologies (FRBRoo 12 , OWL Time 13 , and GeoSPARQL 14 ). NOnt reuses most concepts from the CIDOC CRM ontology - among which is the concept of event - because this is used by many European cultural and museum institutions for data representation. NOnt adds new concepts, such as the role of a narrative actor, the fabula and plot, and new properties, such as the causal dependency between two events and the geospatial belonging of an event to a country or a territory. By reusing concepts from other consolidated ontologies, NOnt enhances interoperability with other vocabularies and semantic knowledge bases and allows for building more extensive knowledge networks in agreement with the Linked Open Data paradigm 15 . From a conservation perspective, our analysed value chains are part of the European cultural heritage and suitably described as narratives because they relate to European territories’ history, artisanal knowledge, and environmental characteristics.

We used narratives for data description also because they are central to human activities across cultural, scientific, and social domains and can establish shared meanings among diverse domain-specific communities 16 . Psychological theories assert that humans comprehend reality by organising events into narratives 17 , 18 . Stories can depict characters’ intentions, emotions, and aspirations through the attributes of objects and events 19 and can describe overall territorial aspects beyond the analytical data 20 , 21 .

Our value chain narratives include comprehensive information on the selected European product value chains, e.g., about the production of European cheese, beer, milk, sheep and dairy farming, flour, herbs, oil, wine, tourism, carpentry, food and drink, nuts, and others. Overall, they cover economic assets, biodiversity, and ecosystem service descriptions (e.g., food and water resources and touristic features). Therefore, our representation also includes geographical information such as maps, pinned locations, and polygonal areas. A map is a valuable support to represent the spatiotemporal structure of a territory story and the relationships between places 21 . For such reason, we also represented the value chain spatiotemporal narratives as Story Maps , i.e., online interactive maps enriched with textual/audio events and digital material narrating overall territorial complexity. Story Maps allow exploring and navigating a narrative through many digital devices (e.g., PCs, tablets, smartphones, interactive displays) and can be built through collaborative tools 20 . They are valuable to represent the life, emotions, reality, fiction, legends, and expectations associated with the described territory beyond a mere map representation 20 , 21 , 22 , 23 , 24 , and fill the perceptual gap between a territory-as-a-whole and its map 25 .

Our principal target data users and stakeholders are policymakers at all spatial scales, from local to European. The knowledge contained in the data is valuable to designing local, regional, national and/or European policies, strategies, and actions to promote the development of mountain areas, starting from the value chains that populate these areas. In fact, the stakeholders can use this knowledge to understand the prevailing economic sector (primary, secondary, or tertiary) of their respective regions. They can also infer information at a finer spatial scale, such as the resources (natural, cultural, and others) on which the productive fabrics depend. Based on this information, they can design place-based and data-driven policies supporting the socio-economic development of marginalised mountain areas. Other stakeholders of our data are citizens who wish to have an overview of their regions’ territorial and economic assets and the related peculiarities, competitions, and challenges in Europe.

Our semantic knowledge graph is available as a collection on a Figshare repository and through a public semantic database instance, and is interactively explorable through online Story Maps ( Data Records and Usage Notes ). To our knowledge, this is the first extensive and structured collection of information on European mountain value chains.

The Figshare collection tries to meet the FAIR (Findable, Accessible, Interoperable, Reusable) principles as far as possible. Figshare indeed fosters the alignment of the hosted data collections to data FAIRness 26 . The data have a unique and persistent Digital Object Identifier (DOI) assigned (Findable-F1 property). The collection’s metadata comply with the DataCite metadata interconnection schema, and we fulfilled all essential and mandatory fields (Findable-F2). The metadata schema contains a dedicated “doi” field (Findable-F3) and is indexed in both the Figshare search engine (without authentication required) and the major search engines (Findable-F4). Moreover, we added textual metadata for each data folder and a data dictionary to improve data interpretability. The data and metadata are accessible for download, without authentication required. They comply with the “Attribution 4.0 International” licence ( CC BY 4.0 ), i.e., they can be copied, redistributed, transformed, and reused even for commercial purposes. Access is also guaranteed through several open authentication protocols (Accessible-A1), and the collection’s metadata and DOI will be preserved for the repository’s lifetime (Accessible-A2). The metadata are accessible through the Figshare APIs and are exportable (through the collection’s page) to several standards (Interoperable-I1). They conform to controlled vocabularies of research categorisation and open licences (Interoperable-I2). The data vocabulary contains a controlled list of concepts belonging to ontological standards (Interoperable-I3). Finally, the metadata description, the open licence, the availability of the input and output data (complemented by provenance description through the present paper), and the use of a semantic knowledge graph for data representation strongly support our collection’s reusability (Reusable-R1 and R2).

Paper and project background

In the present paper, we describe how we built our knowledge graph for 454 European value chains. The primary source data were unstructured textual documents provided by territory experts working in the MOuntain Valorisation through INterconnectedness and Green growth (MOVING) European project 3 . MOVING was an H2020 project (September 2020 - August 2024) involving 23 organizations and companies that monitor, support, and conduct value chains in mountain areas. The primary project target was to collect updated and comparable knowledge on mountainous territories, with the conjecture that this would lead to a deeper understanding of the context, trends, and potential evolution of mountain communities, territories, and businesses. Moreover, this understanding would help design new policies for conservation and evolution. As a main strategy, the project proposed a bottom-up participatory process involving value chain actors, stakeholders, and policymakers to co-design European policy frameworks for resilience and sustainability. The heterogeneous MOVING community of practice monitored 454 value chains. In the first two project years (2020-2021), the territory experts studied and collected local knowledge about geography, traditions, and societal and economic aspects. Each expert independently compiled information on his/her monitored value chains. The provided information was complete from the point of view of the MOVING project scope. The experts used a socio-ecological system (SES) approach to understand the value chain contributions to the mountain areas’ resilience and sustainable development. Within the SES framework, they related the value chain processes and activities to the local natural resources, particularly those affected by climate change and major socioeconomic and demographic trends (e.g., out-migration, livelihoods, and basic-service provisioning). They prioritised land-use and land-use change indicators because most value chains were agri-food and forestry-based, heavily relying on land resources. However, they also included other regional assets when particularly relevant for the region (e.g. hydropower in Scotland, Portugal and Romania; tourism in Italy, Portugal, Spain, Scandinavian countries, Serbia, North Macedonia, Romania, and Bulgaria). The SES approach was also justified by the MOVING project’s focus on understanding the balance between economically valuable activities and environmental protection. Finding the right balance between these contrasting stressors will likely be more difficult in the near future due to the increasing number of European natural protected areas 27 . The possibility of analysing the vulnerabilities of mountainous value chains’ environments, actors, resources, governance, and processes altogether was critical in this context, and could also support decision authorities in the design of multi-actor (public and private) institutional arrangements and multi-level (local, regional, national, and European) policies.

While this approach generated valuable and new knowledge, a side effect was the non-homogeneity of the collected information, e.g., administrative codes and statistical data were sometimes missing, and the investigated territory and value chain data often did not focus on the same assets across the studies. The need for managing this heterogeneous-knowledge scenario was the primary motivation for our study. After approval by the MOVING scientific community, we automatically transformed the unstructured, expert-provided data into a semantic knowledge graph. Here, we also demonstrate - through queries in the SPARQL Protocol and RDF Query Language (SPARQL) - that this representation allows for extracting new valuable knowledge for societal, economic, and environmental monitoring and studies.

The present paper outlines a semi-automated workflow developed to convert unstructured data about European value chains ( VCs ) into a semantic knowledge graph, as depicted in Fig.  1 and elaborated in the current section.

Conceptual flowchart of our data preparation, augmentation, validation, and publication workflow.

Our input was a set of textual documents, each detailing practical aspects of European VCs, including economic, meteorological, climatic, ecological, cultural, and societal aspects, along with specifications about their geographical regions and nations.

During the data preparation phase, these documents were processed to create a preliminary semi-structured form of the VC narratives, organized in tables with rows corresponding to narrative events. Then, a data augmentation phase regarded the extraction of information, for each event, about the mentioned places, locations, organizations, and keywords, and the enrichment of the data with geospatial references. This enriched and structured narrative representation was then converted into a semantic knowledge graph using the OWL format ( Knowledge graph creation and publication ).

This OWL-formatted knowledge graph was subsequently published in an openly accessible online semantic triple store and visually represented through 454 Story Maps. The OWL file, being the main output of this research, is available for other researchers for import into their semantic triple stores ( Usage Notes ). It allows them to explore the rich information about European value chains that the graph encapsulates.

Data preparation

Our data collection originated from textual documents on VCs written by territory experts (researchers, members of local authorities, non-governmental organisations, producers’ and processors’ cooperatives, Local Action Groups, extension services, and others) within the MOVING European project 3 . Each involved country had from 1 to 51 documents associated (Table  1 ). The textual documents coarsely followed one textual-data collection schema designed by researchers at the University of Pisa (UniPi), who were involved in the MOVING project. As a preliminary validation, the UniPi researchers checked each expert’s document for inconsistencies in geographical locations, primary resources, and socioeconomic assets of the reference area and value chain. In the case of inconsistencies identified, they sent the document back to the expert(s) for adjustments, and repeated the checks on the updated document afterwards.

As an additional pre-processing step, we organised the information in the VC documents through an MS Excel table. This table contained one row for each VC and the columns corresponded to different VC aspects (Table  2 ). Some columns contained numeric values (e.g., for incomes and tourism). Other columns contained descriptions in natural language (e.g., the landscape description) or categorised information (e.g., Local Administrative Units). The table was very sparse since information on several columns was often unavailable. This table aimed to provide a first overview of the commonalities and heterogeneity between the VCs across European countries and regions. This file was the only manually prepared dataset of our workflow and the basis of the narrative building and augmentation part described in the next section. The MOVING project experts were also asked to check whether the MS Excel table correctly reported and represented the information they had provided.

As a further pre-processing step, we processed the MS Excel table to produce new tables in Comma Separated Value (CSV) file-format, one for each VC. Each CSV table was a rough, structured version of a VC ( VC table ). Our Figshare repository contains these files for consultation ( Data Records ). Each VC table contained 11 rows corresponding to the key events of a VC narrative (Table  3 -right-hand column). Each row corresponded to one narrative event , with a title and a description column associated. To build the VC tables from the MS Excel table, we implemented a JAVA process that automatically mapped the column contents of one row of the MS Excel table onto the description column of one VC table. Table  3 reports this mapping. For one-to-one mappings, we directly reported the source-column’s text content. When multiple columns corresponded to the same VC event, instead, we appended the column contents through text-harmonisation rules for the conjunctions.

This mapping process produced 454 VC tables, which were the input to the subsequent augmentation phase.

Data augmentation

In the present section, we describe all data augmentation steps in our workflow and eventually report the corresponding algorithm pseudo-codes.

Named entity extraction

Our workflow used a named entity extraction module we implemented in JAVA. This module characterised each event in the VC narrative with abstract or physical objects mentioned in the event description texts ( named entities ). The module used the NLPHub service 28 , a cloud computing service that coordinates and consolidates the results of various state-of-the-art text-mining processes integrated within the D4Science e-Infrastructure 29 , 30 , 31 . In our workflow, we set the NLPHub to identify entities of types location , person , and organisation , plus the keywords of the text. Keywords were individual words or compound terms particularly meaningful within their respective contexts. The NLPHub exploited the D4Science cloud computing platform (named DataMiner 31 ) to efficiently manage the processing of ∼ 5000 event texts in our dataset via distributed and concurrent cloud processing. The named entity extraction module augmented each VC table with one additional column ( named entities ) reporting a comma-separated list of named entities (and keywords) associated with each event.

Wikidata entry association

We used the named entities extracted by the previous step as the input of queries to the Wikidata semantic service’s SPARQL endpoint 32 . A JAVA process executed these SPARQL queries to Wikidata to check if each narrative-event entity could correspond to a Wikidata entry. One special rule was adopted for location -type entities. By convention, Wikidata describes location -type entries with the first letter capitalised. Our process used this convention to check for the existence of Wikidata entries associated with location -type named entities.

In the case of a correspondence found, the process extracted the entry’s Wikidata’s Internationalized Resource Identifier (IRI). The IRI is part of the information the Wikidata SPARQL response returns for an entry. It persists also after entry-content update. For instance, the “Alps” entity had the following Wikidata entry IRI associated: https://www.wikidata.org/wiki/Q1286 which corresponded to the Q1286 identifier.

As an additional step, our process checked the consistency of the Wikidata entry retrieved. In particular, it explored the entry-associated Wikipedia pages. For a Wikidata entry to be valid, its associated Wikipedia pages should not correspond to (i) a disambiguation page, (ii) a page with a title not matching the named entity, or (iii) a page referring to a different named entity type. For example, the Wikipedia page associated with a location -type named entity had to correspond to a location. This check distinguished cases like Tours (the French city) from tours (journeys in the area). These rules overall improved the precision of the association between a Wikidata entry and a named entity, i.e., a validated Wikidata entry likely had the same meaning as the named entity.

At the end of the Wikidata entry retrieval and consistency check, our workflow added one column (named IRIs ) to every VC table. This column contained, for each event, the valid IRIs of the event’s entities. Entities without a valid IRI associated were discarded because they brought the risk of introducing false topics in the narratives.

Geometry association

As an additional data augmentation step, a Python process added a new column (named geometry ) to each VC table containing spatial representations for the location -type entities. The process checked each valid location -type entity for having a corresponding coordinate pair in the associated Wikidata entry. In particular, it retrieved the Wikidata “coordinate location” property (P625) content as a reference longitude-latitude coordinate pair. Moreover, the process also checked if a polygon was possibly associated with the entity. To this aim, it used an instance of the open-access QLever endpoint of the University of Freiburg 33 to retrieve a possible polygon representation from the OpenStreetMap subgraph included in this large knowledge graph. QLever is a SPARQL engine capable of efficiently indexing and querying large knowledge graphs (even with over 100 billion triples) such as Wikidata, Wikimedia Commons, OpenStreetMap, UniProt, PubChem, and DBLP 34 . The University of Freiburg populated a large knowledge graph with these sources. Our process reported all geometries found on the QLever service as Well-Known Text (WKT) formatted strings 35 . The first VC event ( Introduction ), was always assigned the country’s polygon and centroid. Our process added the found entities’ geometries to the geometry column of their associated events. It reported both the polygon and point representations when existing. All geometries reported by our workflow used the EPSG:4326 geodetic coordinate system for World (equirectangular projection).

Representation of Local Administrative Units

The expert-provided data also included the indications of the 2-level Local Administrative Units 36 (LAUs) of the municipalities covered by each VC (Table  2 ). A VC could span more than one municipality and often had several LAUs associated. Eurostat, the statistical office of the European Union, has been producing regional statistics for these areas since 2003 37 . Different LAUs can form one “Nomenclature of Territorial Unit for Statistics” (NUTS), for which Eurostat produces additional statistics. These statistics help assess trends for local community typologies (rural, suburban, and urban), urbanisation degree (city, town and suburb, rural), functional urban areas (cities and their surrounding commuting zones), and coastal areas.

Our workflow included a Python process to retrieve a geometry representation of the VC-associated LAUs (as WKT strings). The process searched for a polygonal representation of each LAU code in two structured files published by Eurostat in their Geographic Information System of the Commission (GISCO)0 38 . GISCO is an EU-funded geographic information system that includes data on administrative boundaries and thematic information (e.g., population data) at the levels of European nations and regions. The first GISCO file our process used was a GeoJSON file 39 containing all WKT polygon representations of the Eurostat-monitored LAUs. However, the experts often reported NUTS codes instead of LAU codes. Therefore, if a polygon representation could not be found for one LAU code, our process searched for the same code in a second GISCO GeoJSON file containing NUTS polygon representations 40 . Since different countries could use the same LAU and NUTS codes for different territories, our process used the VC’s belonging country code (e.g., IT, ES, UK) for disambiguation.

Our process found correspondences for all LAU and NUTS codes associated with our VCs (1224 total). It augmented each VC table’s geometry column with LAU (or NUTS) geometries repeated for each event. It represented all geometries with equirectangular projection, also used in GISCO.

Geometry filtering

The geometries associated with the VC narrative events were checked for “geographical consistency” with the narrative itself. A story map set in the Austrian Alps that mentioned a cow breed also found in America might lead to the inclusion of United States regions’ entities (and thus geometries) in the story. From a narrative point of view, associating a point too distant from the VC territory would be dispersing and produce jittery paths on the map that could confuse the reader. Therefore, we decided to avoid shifts from one continent to another or between far locations in our narratives while keeping a geographically coherent focus.

A dedicated JAVA process estimated a bi-variate log-normal distribution on the longitude-latitude pairs of each narrative. It included the LAU/NUTS centroids among the pairs. The process computed the upper and lower 95% log-normal confidence limits on the coordinates and considered the coordinates outside these boundaries as outliers. Consequently, if most coordinates pertained to a specific region, the calculated boundaries naturally surrounded that region. Otherwise, the boundaries likely encompassed all coordinates if these were uniformly distributed worldwide (e.g., in a global-scale narrative). We demonstrated the validity of a bi-variate log-normal distribution to estimate the primary geographical focus of a narrative in a previous work 20 . Each event in our narrative underwent outlier removal using this log-normal filter. By construction, at least the LAU/NUTS geometries remained associated with an event after the filtering. All geometries associated to an event are reported on a map during the event visualisation in a Story Map.

Image assignment

As a final data augmentation step, our workflow assigned images to the 11 events of each VC narrative through a dedicated Python process. The image associated with the first event ( Introduction ) was always the geographical map of the VC-associated country. This map was retrieved from Wikidata through a SPARQL query on the “locator map image” property (P242). Quantitative events such as “Income and gross value added” and “Employment” were not associated with images because their images should necessarily be conceptual. However, we verified that the MOVING community did not perceive such conceptual images as meaningful. For the remaining events, we used images the MOVING project members willingly provided for each country (without copyright violation). Six images per country were averagely available, which we enriched with additional region-specific images from Wikimedia Commons 41 referring to the VC territories. Our Python process randomly sampled images from the VC’s country-associated image set (without repetitions) and assigned them to the narrative events while prioritising the expert-provided images. For example, the narrative “Chestnut flour coming from the rediscovered chestnut cultivation activities in the area” was enriched with seven images of Tuscany by the MOVING members and two images of the Apuan Alps (where this chestnut flour is produced) from Wikimedia Commons.

In the present section, we report the algorithms of the data augmentation processes described so far.

The data augmentation algorithm for named-entity extraction and geometry association can be summarised as follows:

Algorithm 1

The algorithm translating LAU/NUTS codes into WKT strings is the following:

Algorithm 2

The geometry filtering algorithm can be summarised as follows:

Algorithm 3

The image assignment algorithms can be summarised as follows:

Algorithm 4

Knowledge graph creation and publication

Our workflow used an additional Python process to translate all augmented VC tables into a semantic knowledge graph. As a first step, this process translated the VC tables into JSON files that followed a schema we designed and optimised in a previous work 20 , 42 . This JSON representation structurally describes the event sequence and the associated entities, images, geometries, and Wikidata IRIs. Our process also stored each JSON file in a PostgreSQL-JSON database for quick retrieval and use for narrative visualisation ( Usage Notes ).

As a second step, the process translated each JSON file into a Web Ontology Language (OWL) graph file and assembled all graphs into one overall OWL graph file 43 . To this aim, it invoked a JAVA-based semantic triplifier software we implemented for this specific sub-task. The VC-individual and the overall OWL graphs complied with the OWL 2 Description Logic 44 , which assured the decidability of the language. They adhered to the Narrative Ontology model version 2.0 7 , 45 , extended with the GeoSPARQL ontology 14 , a standard of the Open Geospatial Consortium that handles geospatially explicit data in ontologies. We published the entire VC-narrative OWL graph (and sub-graphs) in the Figshare repository attached to the present paper ( Data Records ) to openly allow users to import them in a semantic triple store and query, explore, and infer knowledge on the 454 European VCs represented. This file was the main output of our workflow.

We also published the knowledge graph on a public-access Apache Jena GeoSPARQL Fuseki 46 semantic triple store to allow users to execute semantic and geospatial queries to our knowledge graph openly ( Usage Notes ). Fuseki is a SPARQL service that can internally store Resource Description Framework (RDF) data representing semantic triples consisting of a subject (node), a predicate (relation) and an object (node) of the knowledge graph. This service allows retrieving the stored RDF triples through a SPARQL/GeoSPARQL endpoint.

It is important to stress that the main target of the knowledge graph was to enhance the communication about the value chains to a broad, heterogeneous audience. Our target stakeholders were value chain and territory experts, citizens, and local and national administrations. These stakeholders need an overall understanding of the value chains, their role in characterising the territory, and the criticalities and strengths of the territory. Entities such as locations, persons, organisations, and keywords - enriched with images and geometries - matched their interests. Consequently, we did not include statistics and numeric indicators among the entities because data analytics was not a target of the narratives. Moreover, the unavailability of statistical data for several value chains would have created knowledge gaps across the narratives. Therefore, we reported statistical data, when available, in an understandable and readable textual format in the event text while leaving the possibility to conduct data analytics on the tabular-format files available in the Figshare repository.

As an additional note, we clarify that semantic knowledge graphs were a more suitable choice for data representation than tabular data models. Tabular data models, such as those used in relational databases, satisfy a predefined schema. Coercing our data to rows and columns was unsuitable for quickly capturing the complex relationships between value chain entities. Moreover, tabular models hardly manage inconsistent naming conventions, formats, and ambiguous identifiers like those in our data. Although foreign keys allow for modelling rich, interconnected data, they introduce complexity in the database schema, making knowledge extraction more challenging. Moreover, as the volume of data grows, managing and querying large relational tables can become inefficient and require dedicated distributed systems. In a scenario like ours, where data were many, heterogeneous, and dynamic, we could not assume that a traditional relational schema was efficient and effective. Instead, we used Linked Data and Semantic Web technologies because they offered more flexibility in quickly extending, modifying, and interconnecting a knowledge base of diverse and heterogeneous data. Moreover, semantic graphs could intuitively represent rich and complex relationships between the data while capturing real-world facts. They also enacted interoperability through the reuse of shared vocabularies and IRIs from other semantic knowledge bases, allowing the creation of interconnected, consistent data networks. Finally, as semantic technologies are Web-native, they quickly allowed for accessing and querying data through standard Web protocols.

Data Records

We made the data available on a public-access Figshare repository 47 (version 3, currently). One dataset is available for downloading the overall OWL knowledge graph, which allows other users to reproduce the entire knowledge base in another triple store ( Usage Notes ). This knowledge graph contains 503,963 triples. The data in the graph are also available in CSV and GeoPackage formats for easy import, manipulation, and inspection in multiple software solutions. Another dataset presents a folder hierarchy containing sub-graphs focusing on one VC at a time. The folder organisation is optimised for regional ecological, socioeconomic, and agricultural modelling experts. The files are organised into subfolders, each corresponding to a country. The name of each file reports the title of the corresponding value chain. Each file is in the OWL format (e.g. wood_charcoal_from_Gran_Canaria_island.owl). The complete file collection contains 454 files that can be imported independently of each other. The Figshare repository also contains all links to the JAVA and Python software used in our workflow.

Additionally, the repository contains a direct link to our public-access Apache Jena GeoSPARQL Fuseki instance hosting the entire VC knowledge graph. This instance allows the execution of SPARQL and GeoSPARQL queries ( Usage Notes ). The Figshare repository also contains the MS Excel file that was the input of our workflow. It allows for comparing our workflow’s original and final products and repeating the technical validation. Finally, the repository contains all VC tables in CSV format resulting from the data preparation phase. The authors received authorisation by the MOVING project community to publish this material.

Technical Validation

Formal consistency of the knowledge graph.

We used a semantic reasoner to validate the logic consistency of our entire OWL graph. A semantic reasoner is software designed to infer logical consequences from a set of asserted facts or axioms. We used the Openllet open-source semantic reasoner 48 , 49 , 50 to (i) check the consistency of our OWL graph (i.e., to guarantee that it did not imply contradictions), (ii) check that the class hierarchy respected the one of the Narrative Ontology, (iii) test geometry consistency (polygon closures and correct WKT formatting), (iv) test the possibility to execute complex SPARQL and GeoSPARQL queries.

Openllet assessed the consistency of our knowledge graph on all the checks reported above. The reasoner confirmed that the subclass relations and the complete hierarchy between the classes fully respected the ones of the Narrative Ontology. The class hierarchy allowed the correct extraction of all subclasses of a class. Finally, all geometries were assessed as consistent. GeoSPARQL queries allowed the execution of spatial reasoning and all algebraic operations between sample-selected polygonal geometries from the VCs.

Additionally, we executed automatic checks to ensure that no event in the OWL graph was empty or contained meaningless or misreported content (e.g., “N/A”, “Empty”, “Unavailable”, etc.). The checks also verified that every LAU had an associated WKT polygon from the GISCO database, and we manually verified that the correspondences were correct. An expert from the MOVING project also conducted a sample check to verify that the mapping between the pre-processed MS Excel table columns and the VC-narrative events (Table  3 ) produced meaningful and human-readable descriptions.

Performance of the named entity extraction-filtering process

We evaluated the performance of the combined process of named entity extraction plus Wikidata entry association (filtering). This process influences the results of the queries to the knowledge graph. An incomplete set of entities would indeed limit the information associated with an event. Moreover, as highlighted in the next section, it would also limit the discovery of interconnections between the events. When querying for events with several associated entities, a target event would only be selected if all entities were retrieved correctly.

To measure the quality of the named entity extraction-filtering process, we evaluated its information-extraction performance on manually annotated story events. To this aim, we selected a statistically meaningful set of events to annotate through the sample size determination formula with finite population correction , i.e.

where n is the target sample size adjusted for a finite population; n 0 is the initial sample size assuming an infinite population; Z is the Z-score corresponding to the desired confidence level (Z-score =1.96 for the 95% confidence level we used); p is the prior assessment (0.5 for uninformative conditions); MOE is the margin of error on the true error (5%), and N is the total number of events (population size).

This formula estimated that 330 events (corresponding to 30 stories) could be sufficient for performance assessment. Consequently, we randomly selected 30 stories from our collection. In this sub-collection, we identified the automatically extracted-filtered entities correctly associated with key event-related concepts ( true positives , TP). Then, we identified those unrelated to key concepts ( false positives , FP). Finally, we annotated additional entities (with valid Wikidata pages associated) from the events’ texts that the extraction-filtering process missed ( false negatives , FN). Based on these annotations, we calculated the following standard performance measurements:

The evaluation results are reported in Table  4 . The high Precision (0.99) suggests that most extracted entities were correctly associable with key concepts expressed in the events. Maximising the reliability of the extracted entities was indeed the principal target of our entity extraction-filtering process. Instead, the lower Recall (0.93) suggests that the extracted entity set could be incomplete and could negatively impact multi-entity querying. However, the F1 measure (0.96) was one of a good information extraction system. Therefore, although the extraction-filtering process could be improved (which will be part of our future work), its performance was sufficiently high to support our knowledge graph reasonably.

Query-based validation

We verified that our knowledge graph could contribute to discovering new knowledge from the data, which was its principal aim. In particular, we collected the types of interrogations the MOVING-project scientists or stakeholders considered valuable, i.e., hard to identify without a semantic knowledge representation. These interrogations were collected (i) during plenary meetings with the MOVING community of practice, (ii) after identifying the principal study targets of the rural-area experts involved in the project (typically value chains within their territories), and (iii) by reading project deliverables. For example, the experts’ targets were the VCs sharing common environmental characteristics (e.g. rivers, lakes, vineyards, and chestnut trees), issues (e.g. depopulation, pollution, and deforestation), and similar products (e.g. cow/sheep milk and cheese). Discovering this knowledge from the data holds significant value for mountain ecosystems, as it aids in planning sustainable environmental management strategies 21 . Additionally, this knowledge is valuable in supporting the long-term ecological sustainability of urban areas and comprehending and mitigating the decline of fundamental services in mountain areas brought about by the ongoing depopulation trends 2 , 51 , 52 . We demonstrated that our knowledge graph could contribute to these directions.

We focussed on ten types of knowledge-extraction targets, corresponding to ten SPARQL/GeoSPARQL queries regarding different and complementary aspects of European mountain products and their related spatial distributions. In particular, we extracted the VCs with the following characteristics:

related to vineyard products (Q1)

possibly affected by deforestation (Q2)

involving cheese with Protected Designation of Origin (PDO) certification (Q3)

producing cheese made with cow and/or goat milk (Q4)

using sheep to produce cheese (Q5)

using sheep to produce wool (Q6)

operating in the Alps (Q7)

operating around Aosta city (Italy) (Q8)

operating in Scotland (Q9)

operating around long Italian rivers (>100) (Q10)

The information extracted by these queries overall covered the interests of the MOVING community experts. It would have been hard, indeed, to extract the same information through the usual data representation and technology adopted by this scientific community. Based on the query results, we calculated Precision, Recall, and F1. We showed that high Precision was achieved for most cases, i.e., even when the information retrieved was incomplete (mostly due to misdetection by the named entity extraction processes) the results were reliable. The performance measurements (Table  5 ) demonstrate the overall high quality of our knowledge graph and the general effectiveness of the queries. In the following, we report the details of the queries and the corresponding results.

Q1 - Value chains related to vineyard products

In the following, we report the SPARQL query corresponding to Q1.

SPARQL Query 1

PREFIX narra: < https://dlnarratives.eu/ontology# >

PREFIX ecrm: < http://erlangen-crm.org/current/ >

PREFIX rdfs: < http://www.w3.org/2000/01/rdf-schema# >

SELECT DISTINCT?title?country WHERE { ?event1 narra:partOfNarrative?narrative. ?narrative rdfs:label?title. ?narrative narra:isAboutCountry? countryIRI. ?countryIRI rdfs:label?country. {

?event1 narra:hasEntity < https://dlnarratives.eu/resource/Q22715 >.

?event1 narra:hasEntity < https://dlnarratives.eu/resource/Q282 >.

?event1 narra:hasEntity < https://dlnarratives.eu/resource/Q10978 >.

?event2 narra:partOfNarrative?narrative.

?event2 narra:hasEntity < https://dlnarratives.eu/resource/Q282 >.

FILTER (?event1! = ?event2) } UNION {

?event2 narra:partOfNarrative?narrative .

?event2 narra:hasEntity < https://dlnarratives.eu/resource/Q10978 >.

FILTER (?event1 ! = ? event2)

?event1 narra:hasEntity <https://dlnarratives.eu/resource/Q22715 >.

?event3 narra:partOfNarrative?narrative.

?event3 narra:hasEntity < https://dlnarratives.eu/resource/Q10978 >.

FILTER (?event1! = ?event2 & &?event1! = ?event3 & &?event2! = ?event3) } } ORDER BY lcase(?country)

The query retrieves distinct titles of narratives along with their associated countries. Several ontology prefixes are specified at the beginning of the query to shorten the corresponding IRIs, which are used in the subsequent parts of the query. The SELECT statement specifies the variables (“?title” and “?country”) whose values the query will output. The WHERE clause contains the conditions that need to be satisfied for each result. It involves several semantic-triple patterns connected by the “.” operator:

The first triple pattern (“?event1 narra:partOfNarrative?narrative”) connects an event to a narrative;

The second triple pattern (“?narrative rdfs:label?title”) retrieves the label (title) of the narrative.

The third triple pattern (“?narrative narra:isAboutCountry?countryIRI”) connects the narrative to its related country.

The fourth triple pattern (“?countryIRI rdfs:label?country”) retrieves the label (name) of the country.

The subsequent UNION blocks combine pattern alternatives, each representing a condition under which events are selected. The blocks retrieve events associated with at least one entity among “vineyard” (id. Q22715), “wine” (id. Q282), and “grape” (id. Q10978). These entities were chosen with the help of an expert. They are the entities most related to vineyards in our knowledge graph. The expert was aided by an entity search tool included in our visualisation facilities ( Usage Notes ).

The sets of narrative events containing the entities above are labelled “event1”, “event2”, and “event3”, respectively. Filters are applied (e.g., “FILTER (?event1! = ?event2)”) to ensure that the entities can singularly appear in different events.

Finally, the ORDER BY clause sorts the results alphabetically by the lowercase label of the country. In the case of multiple sub-graphs imported instead of the overall graph, the query should be changed by adding a “FROM <urn:x-arq:UnionGraph>” clause, before the WHERE clause, to specify that the query should be conducted on the union of the sub-graphs.

In summary, this query retrieves the titles of the narratives and their associated countries, comprising events related to the “vineyard”, “wine”, and “grape” entities. The query produced the output reported in Table  6 . To verify the correctness of the retrieved information, we manually checked, with the help of a MOVING expert, the VCs (among the 454) that contained information on vineyard products. Precision and Recall (0.93 and 0.90, respectively), were reasonably high, and F1 (0.91) indicated good retrieval performance. The main reason for Recall not reaching one was the presence of faults (false negatives) by the named entity extraction processes in detecting vineyard-related entities in the event texts. Precision was, instead, negatively affected by citations of vineyard-related products in VCs that did not focus on vineyard products (false positives).

Q2 - Value chains possibly affected by deforestation

In the following, we report the SPARQL query corresponding to Q2.

SPARQL Query 2

SELECT DISTINCT?title?country WHERE {?event narra:partOfNarrative?narrative. ?narrative narra:isAboutCountry? countryIRI. ?countryIRI rdfs:label?country. ?narrative rdfs:label?title.

?event narra:hasEntity < https://dlnarratives.eu/resource/Q169940 >.

} order by lcase(?country)

Similarly to Q1, this query retrieves the titles and associated countries of the narratives mentioning deforestation. The notable difference compared to Q1 is in the WHERE clause, which retrieves the events (“?event narra:hasEntity < https://dlnarratives.eu/resource/Q169940 >”) having “deforestation” (id. Q169940) among the associated entities.

This query produced the result reported in Table  7 . Expert verification assessed that it retrieved the complete and correct set (Precision and Recall equal to 1) of all VCs affected by deforestation. Therefore, this query shows the value of our knowledge graph for discovering critical threats to the VCs and their territories.

Q3 - Value chains involving cheese with Protected Designation of Origin certification

In the following, we report the SPARQL query corresponding to Q3.

SPARQL Query 3

PREFIX narra: < https://dlnarratives.eu/ontology#>

SELECT DISTINCT?title?country WHERE { ?event1 narra:partOfNarrative?narrative. ?narrative rdfs:label?title. ?narrative narra:isAboutCountry?countryIRI. ?countryIRI rdfs:label?country. {

?event1 narra:hasEntity < https://dlnarratives.eu/resource/Q10943 >.

?event2 narra:hasEntity < https://dlnarratives.eu/resource/Q13439060 >.

FILTER (?event1! = ?event2) } } ORDER BY lcase(?country)

The query structure is similar to the one of Q1, with the difference that the entities “cheese” (id. Q10943) and “Protected designation of origin” (PDO) (id. Q13439060) are used to detect events, and consequently, the target VCs.

The query produced the results reported in Table  8 . This case is peculiar because it demonstrates the potential bottleneck of the performance of the named entity extraction processes. Although the query did not produce false positives (i.e., Precision was 1), there were many false negatives due to frequently missed recognition of PDO mentions in the event texts (Recall was 0.26). One reason is that long and articulated entities like “Protected designation of origin” are often subject to misspelling, abbreviation, and native-language reporting (e.g., DOP, in Italian), which prevent algorithms from identifying them. Therefore, Q3 showed a potential limitation of our knowledge graph when searching articulated entities. However, the same complexity of these entities ensured that the results were correct when the entities were identified.

Q4 - Value chains producing cheese made with cow and/or goat milk

In the following, we report the SPARQL query corresponding to Q4.

SPARQL Query 4

?event2 narra:hasEntity < https://dlnarratives.eu/resource/Q2934 >.

?event1 narra:hasEntity < https://dlnarratives.eu/resource/Q830 >.

?event1 narra:hasEntity < https://dlnarratives.eu/resource/Q11748378 >.

?event2 narra:hasEntity < https://dlnarratives.eu/resource/Q11748378 >.

?event2 narra:hasEntity < https://dlnarratives.eu/resource/Q830 >.

?event1 narra:hasEntity < https://dlnarratives.eu/resource/Q10943> .

?event3 narra:hasEntity < https://dlnarratives.eu/resource/Q2934 >.

?event4 narra:partOfNarrative?narrative .

?event4 narra:hasEntity < https://dlnarratives.eu/resource/Q830 >.

FILTER (?event1! = ?event2 && ?event1! = ?event3 && ?event2! = ?event3 && ?event1! = ?event4 && ?event2! = ?event4 && ?event3! = ?event4) } } ORDER BY lcase(?country)

This query operates a search for narratives bound on four entities: “cheese” (id. Q10943), “cow” (id. Q11748378), “goat” (id. Q2934), and “cattle” (id. Q830). The query structure is similar to Q1.

The query produced the results reported in Table  9 . The results were still affected by the named entity extraction bottleneck because the query’s success depended on the correct identification of all four terms in a narrative. Compared to Q3, the present query tested the retrieval of multiple, simpler terms. Precision (0.84) and Recall (0.55) were indeed higher than the one of the articulated-entity search of Q3 (0.26).

Q5-Q6 - Value chains using sheep to produce cheese vs wool

In the following, we report the SPARQL queries corresponding to Q5 and Q6.

SPARQL Query 5 - VCs using sheep to produce cheese

SELECT DISTINCT?title?country WHERE { ?event1 narra:partOfNarrative?narrative. ?narrative rdfs:label?title. ?narrative narra: isAboutCountry?countryIRI. ?countryIRI rdfs:label?country. {

?event1 narra:hasEntity < https://dlnarratives.eu/resource/Q7368 >.

?event2 narra:hasEntity < https://dlnarratives.eu/resource/Q10943 >.

FILTER (?event1! = ?event2)

ORDER BY lcase(?country)

SPARQL Query 6 - VCs using sheep to produce wool

SELECT DISTINCT?title?country WHERE { ?event1 narra:partOfNarrative?narrative. ?narrative rdfs:label?title. ?narrative narra:isAboutCountry?countryIRI. ?countryIRI rdfs:label?country.

{?event1 narra:hasEntity < https://dlnarratives.eu/resource/Q7368 >.

?event2 narra:hasEntity < https://dlnarratives.eu/resource/Q42329 >.

These queries have the same structure as Q1. They share one entity, “sheep” (id. Q7368), with two different usages (corresponding to different entities used in the WHERE clause), i.e., “cheese” (id. Q10943) in Q5 and “wool” (id. Q42329) in Q6.

The results are reported in Tables  10 and 11 . The performance measurements between the two queries were very similar. The false positives, which affected Precision, were due to mentions of sheep in other VCs that did not regard usage for cheese or wool production. Notably, although the two queries retrieved mostly different VCs, the fraction of correct narratives retrieved (Precision) was 0.69 in each case. Moreover, Recall values (0.72 and 0.75, respectively) were similar, due to similar fractions of undetected mentions (false negatives) of cheese and wool by the named entity extraction processes. Overall, a ∼ 0.70 F1 for both queries indicated an overall moderate-high reliability of the results.

Q7 - Value chains operating in the Alps

In the following, we report the GeoSPARQL query corresponding to Q7.

GeoSPARQL Query 7

PREFIX geof: < http://www.opengis.net/def/function/geosparql/ >

PREFIX geo: < http://www.opengis.net/ont/geosparql# >

PREFIX osm: < https://www.openstreetmap.org/ >

PREFIX wd: < http://www.wikidata.org/entity/ >

PREFIX osm2rdfkey: < https://osm2rdf.cs.uni-freiburg.de/rdf/key# >

SELECT?nlabel?clabel?wktLau WHERE { ?narra narra:isAboutCountry?country; narra:isAboutLAU?lau; rdfs:label?nlabel. ?country rdfs:label?clabel. ?lau geo:hasGeometry?glau. ?glau geo:asWKT?wktLau. { SELECT?wkt WHERE {

< https://qlever.cs.uni-freiburg.de/api/osm-planet > {

?osm_id osm2rdfkey:wikidata wd:Q1286; geo:hasGeometry?geometry. ?geometry geo:asWKT?wkt. } } } FILTER(geof:sfIntersects(?wktLau,?wkt)). }

The query retrieves the VC narrative titles, countries, and LAU polygons that overlap a polygon defining the Alps region. A value chain’s LAUs define the main areas where the VC operates (i.e., produces and sells products). The query internally calls the QLever endpoint provided by the University of Freiburg (Section Geometry association ), and in particular, the Open Street Map (“oms”) subgraph, to define the Alps polygonal region. The SELECT statement specifies the variables “nlabel” (narrative title), “clabel” (country name) and “wktLau” (LAU geometry in WKT format) that will be the output of the query. The WHERE clause contains the conditions that should be satisfied by each result. Differently from the previous queries, the following patterns are included:

The triple pattern “?narrative narra:isAboutLAU?lau” connects a narrative to the corresponding LAU;

the triple pattern “?lau geo:hasGeometry?glau” retrieves the geometry of the LAU;

the triple pattern “?glau geo:asWKT?wktLau” retrieves the WKT description of the LAU geometry;

A nested SELECT clause retrieves the WKT description (“?wkt”) under the following WHERE conditions:

The SERVICE keyword is used to invoke the external QLever endpoint (“ https://qlever.cs.uni-freiburg.de/api/osm-planet ”);

The triple pattern “?osm_id osm2rdfkey:wikidata wd:Q1286” retrieves the instance corresponding to the QLever entity “Alps” (wd:Q1286);

The triple pattern “?osm_id geo: geometry?geometry” retrieves the geometry-object of “Alps”;

The triple pattern “?geometry geo:asWKT?wkt” retrieves the WKT format of the “Alps” geometry.

A final FILTER clause operates the intersection between the LAU and the “Alps” geometries and retrieves all LAU geometries intersecting “Alps”. The set of LAUs returned by this query can be imported into a Geographic information system (GIS) visualiser and overlapped with the reference region (Fig.  2 ).

Comparison between the Alps regions (as a red polygon) and the Local Administrative Units (orange polygons) of the value chains operating in this region. An interactive map is also reported in the Figshare repository associated with the present article.

The expert’s evaluation highlighted that the LAUs this query retrieved were correct and complete (Precision and Recall were 1). Therefore, the query was valuable in retrieving region-specific VCs.

Q8 - Value chains operating around Aosta city (Italy)

In the following, we report the GeoSPARQL query corresponding to Q8.

GeoSPARQL Query 8

PREFIX uom: < http://www.opengis.net/def/uom/OGC/1.0/ >

SELECT?nlabel?clabel?wktLau WHERE { { ?narra narra:isAboutCountry?country; narra:isAboutLAU?lau; rdfs:label?nlabel. ?country rdfs:label?clabel. ?lau geo:hasGeometry?glau. ?glau geo:asWKT?wktLau. } FILTER(geof:sfIntersects( ?wktLau, geof:buffer( "POINT(7.3196649 45.7370885)"^^geo:wktLiteral, 0.3, uom:degree))). }

This query extracts the VC titles, countries and LAU geometries of the value chains operating within a maximum distance of 23 km from Aosta. The query structure is similar to the one of Q7, with the difference that it does not use an external endpoint to retrieve the reference geometry. Instead, the FILTER clause operates an intersection between all VCs’ LAU geometries and a circular buffer of 0.3 degrees ( ∼ 23 km) around the Aosta longitude-latitude coordinates.

The query produced the results visualised in Fig.  3 . As in the case of Q7, the expert’s evaluation highlighted that the LAUs retrieved by this query were correct and complete (Precision and Recall were 1). Therefore, the query was valuable in retrieving city-specific VCs.

Highlight of the Local Administrative Units (orange) of the value chains operating in a circle denoting a 0.3-degree area around Aosta, Italy (red). An interactive map is also reported in the Figshare repository associated with the present article.

Q9 - Value chains operating in Scotland

In the following, we report the GeoSPARQL query corresponding to Q9.

GeoSPARQL Query 9

PREFIX osmrel: < https://www.openstreetmap.org/relation/ >

PREFIX schema: < http://schema.org/ >

SELECT?nlabel?clabel?wktLau WHERE { ?narra narra:isAboutCountry?country; narra:isAboutLAU?lau; rdfs:label?nlabel. ?country rdfs:label?clabel. ?lau geo:hasGeometry?glau. ?glau geo:asWKT?wktLau. {SELECT?wkt WHERE {

< https://qlever.cs.uni-freiburg.de/api/osm-planet{

?osm_id osm2rdfkey:wikidata wd:Q22;

a osm:relation; geo:hasGeometry?geometry. ?geometry geo:asWKT?wkt. } } } FILTER(geof:sfWithin(?wktLau,?wkt)). }

This query extracts the VC titles, countries, and LAU geometries of the value chains operating in Scotland. The query structure is still similar to that of Q7. It uses the same external QLever Open Street Map endpoint to retrieve the geometry of Scotland boundaries. The FILTER clause operates the intersection between Scotland and the VCs’ LAU geometries.

The query produced the results reported in Fig.  4 . The expert’s evaluation highlighted that the LAUs this query retrieved were correct and complete (Precision and Recall were 1). Therefore, the query was valuable in retrieving country-specific VCs.

Highlight of the Local Administrative Units (orange) of the value chains operating in a polygon defining Scotland national boundaries (red). An interactive map is also reported in the Figshare repository associated with the present article.

Q10 - Value chains operating around long Italian rivers

In the following, we report the GeoSPARQL query corresponding to Q10.

GeoSPARQL Query 10

PREFIX osmkey: < https://www.openstreetmap.org/wiki/Key: >

PREFIX wdt: < http://www.wikidata.org/prop/direct/ >

SELECT?nlabel?clabel?wktLau WHERE { ?narra narra:isAboutCountry?country; narra:isAboutLAU?lau; rdfs:label?nlabel. ?country rdfs:label?clabel. ?lau geo:hasGeometry?glau. ?glau geo:asWKT?wktLau. { SELECT?river_osm?river_wd?river_name?length?wkt WHERE {

SERVICE < https://qlever.cs.uni-freiburg.de/api/osm-planet>{

?river_osm a osm:relation;

osmkey:waterway?waterway; geo:hasGeometry?geometry; osmkey:name?river_name; osm2rdfkey:wikidata?river_wd. ?geometry geo:asWKT?wkt.

SERVICE < https://qlever.cs.uni-freiburg.de/api/wikidata > {

?river_wd wdt:P31/wdt:P279* wd:Q4022; wdt:P17 wd:Q38; wdt:P2043?length.

FILTER (?length > 100)

} } } ORDER BY DESC(?length) }

FILTER(geof:sfIntersects(?wktLau,?wkt)) .

This query retrieves all VCs operating close to an Italian river longer than 100 km. The query structure is very similar to that of Q7 and uses the same external endpoint. The main differences are the following:

the nested SELECT clause operates on two different QLever-instance subgraphs: Open Street Map and Wikidata. The query retrieves the river geometries from the first. Then it uses the second to retrieve the list of “rivers” (id. Q4022) present in “Italy” (id. Q38) whose “length” (id. P2043) exceeds 100 km (“FILTER (?length > 100)”);

the final FILTER clause operates the intersection between the Italian rivers and the VCs’ LAU geometries.

The query produced the results reported in Fig.  5 . All VCs retrieved were correct and complete (Precision and Recall were 1). Therefore, the query was valuable in retrieving river-related VCs and, by extension, could be used to extract water-basin-related VCs.

Highlight of the Local Administrative Units (orange) of the value chains intersecting Italian rivers longer than 100 km (red). An interactive map is also reported in the Figshare repository associated with the present article.

Usage Notes

Our open-access Figshare repository 47 (version 3, currently) contains the entire OWL file and 454 OWL files corresponding to all VC tables. To perform non-explicit geospatial queries, a user can download and import the whole OWL graph (or a subset of the 454 OWL files to focus on specific regions or value chains) in an Apache Jena Fuseki triple store instance on a local machine 46 . After downloading and installing Fuseki 53 , users should access the server interface through a Web browser and navigate to the “manage” interface section. Then, by clicking on the “new dataset” tab, they should create a new dataset, specifying a name and type (i.e., In-memory or Persistent). Next, they should select the just-created dataset and upload the entire OWL file (or a subset of the 454 OWL files) by clicking the “add data” button. Users can verify if the dataset was successfully populated by executing queries on the Fuseki SPARQL query interface. To perform geospatially explicit queries, a user should download and install Apache Jena GeoSPARQL Fuseki 54 , a Fuseki version enhanced with GeoSPARQL features. Currently, this version does not have a Web interface to facilitate data import. Therefore, users should import data programmatically through the GeoSPARQL module embedded with this service 54 . The entire geospatialised narrative events’ data (text, entities, geometries) in the knowledge graph are also available in CSV and GeoPackage formats, which can be imported, visualised, and manipulated through GIS software (e.g., QGIS 55 ).

Our Figshare repository also contains a link to an overall visualisation facility for the VC narratives in the form of interactive and navigable Story Maps 56 . This facility allows our final users and stakeholders to easily explore the value chain locations, entities, events, and images. An overall map shows the distribution of the 454 VC narratives. After clicking on one reference pin, the user is redirected to the Story Map of the corresponding VC narrative. The Story Map lets the user go through the story events while panning, zooming, and inspecting the map locations at the right-hand side of the visualisation panel. The Story Maps are also available as an offline-visualisable HTML page collection in the Figshare repository.

Each Story Map allows users to access a “Search” Web interface (through a button at the top-left of the introductory page) that offers a visual search tool that executes semantic queries behind the scenes. This functionality interrogates the entire knowledge graph residing on a public-access Apache Jena GeoSPARQL Fuseki instance we currently offer and maintain 57 . The “Search” interface allows users to augment the knowledge reported in one Story Map through the knowledge in all other narratives. The “Search” functionality uses predefined SPARQL queries to extract:

All stories in which an entity appears;

All events across all narratives in which an entity appears;

The number of occurrences of one entity across all narratives;

The entities that co-occur with one entity across all events of all narratives.

The purpose of this feature is to allow all users (also without skills in formal semantics) to explore narrative interconnections.

Code availability

Our public-access Figshare repository 47 contains all the JAVA and Python programs we used to execute the described workflow, along with the data inputs and outputs. It allows other scientists to conduct technical validation. The repository also contains an interactive, offline-visualisable HTML version of the result tables and maps.

Our code and data also embed information from the following external sources: Wikidata 58 , OpenStreetMap 59 through the open-access QLever endpoint of the University of Freiburg 33 , and geographic area definitions from Eurostat-GISCO 38 , 39 , 40 .

UNPD (United Nations, D. o. E. & Social Affairs, P. D. World urbanization prospects: The 2018 revision (st/esa/ser.a/420). https://population.un.org/wup/Publications/Files/WUP2018-Report.pdf (2018).

United Nations. Sustainable Development Goal 11, “Make cities and human settlements inclusive, safe, resilient and sustainable”. on-line https://sdgs.un.org/goals/goal11 Accessed 4 January 2023 (2015).

H. Moving (mountain valorisation through interconnectedness and green growth https://www.moving-h2020.eu/ (2020).

Lehmann, F. Semantic networks. Computers & Mathematics with Applications 23 , 1–50 (1992).

Article   MathSciNet   CAS   Google Scholar  

Hogan, A. et al . Knowledge graphs. ACM Computing Surveys (Csur) 54 , 1–37 (2021).

Article   Google Scholar  

Guarino, N., Oberle, D. & Staab, S. What is an ontology? Handbook on ontologies 1–17 (2009).

Meghini, C., Bartalesi, V. & Metilli, D. Representing narratives in digital libraries: The narrative ontology. Semantic Web 12 , 241–264 (2021).

CRAEFT European Project. Craft Understanding, Education, Training, and Preservation for Posterity and Prosperity (CRAEFT) Project Web site. on-line Accessed 12 July 2024 https://www.craeft.eu/ (2024).

Mingei European Project. Mingei - Representation and Preservation of Heritage Crafts Project Web site. on-line Accessed 12 July https://www.mingei-project.eu/ (2024).

IMAGO Italian PRIN Project. Index Medii Aevi Geographiae Operum (IMAGO) Project Web site. on-line https://imagoarchive.it Accessed 12 July (2024).

Doerr, M. The cidoc conceptual reference module: an ontological approach to semantic interoperability of metadata. AI magazine 24 , 75–75 (2003).

Google Scholar  

Bekiari, C. et al . Definition of FRBRoo: A conceptual model for bibliographic information in object-oriented formalism. International Federation of Library Associations and Institutions (IFLA) repository https://repository.ifla.org/handle/123456789/659 (2017).

Pan, F. & Hobbs, J. R. Time ontology in owl. W3C working draft, W3C 1 , 1 (2006).

Battle, R. & Kolas, D. Geosparql, enabling a geospatial semantic web. Semantic Web Journal 3 , 355–370 (2011).

Thanos, C., Meghini, C., Bartalesi, V. & Coro, G. An exploratory approach to data driven knowledge creation. Journal of Big Data 10 , 1–15 (2023).

McInerny, G. J. et al . Information visualisation for science and policy: engaging users and avoiding bias. Trends in ecology & evolution 29 , 148–157 (2014).

Bruner, J. The narrative construction of reality. Critical inquiry 18 , 1–21 (1991).

Taylor, C. Sources of the self: The making of the modern identity (Harvard University Press, 1992).

Delafield-Butt, J. T. & Trevarthen, C. The ontogenesis of narrative: from moving to meaning. Frontiers in psychology 6 , 1157 (2015).

Article   PubMed   PubMed Central   Google Scholar  

Bartalesi, V., Coro, G., Lenzi, E., Pagano, P. & Pratelli, N. From unstructured texts to semantic story maps. International Journal of Digital Earth 16 , 234–250 (2023).

Article   ADS   Google Scholar  

Bartalesi, V. et al . Using semantic story maps to describe a territory beyond its map. Semantic web (Online) 1–18, https://doi.org/10.3233/SW-233485 (2023).

Caquard, S. & Cartwright, W. Narrative cartography: From mapping stories to the narrative of maps and mapping. The Cartographic Journal 51 , 101–106, https://doi.org/10.1179/0008704114Z.000000000130 (2014).

Peterle, G. Carto-fiction: narrativising maps through creative writing. Social & Cultural Geography 20 , 1070–1093, https://doi.org/10.1080/14649365.2018.1428820 (2019).

Bartalesi, V., Metilli, D., Pratelli, N. & Pontari, P. Towards a knowledge base of medieval and renaissance geographical latin works: The imago ontology. Digital Scholarship in the Humanities https://doi.org/10.1093/llc/fqab060 (2021).

Korzybski, A. A non-aristotelian system and its necessity for rigour in mathematics and physics. In Science and sanity: an introduction to non-Aristotelian systems and general semantics (Lancaster, 1933).

Figshare. How Figshare aligns with the FAIR principles. Figshare Web site  =  https://help.figshare.com/article/how-figshare-aligns-with-the-fair-principles (2024).

European Environment Agency. Terrestrial protected areas in Europe. https://www.eea.europa.eu/en/analysis/indicators/terrestrial-protected-areas-in-europe (2023).

Coro, G., Panichi, G., Pagano, P. & Perrone, E. Nlphub: An e-infrastructure-based text mining hub. Concurrency and Computation: Practice and Experience 33 , e5986 (2021).

Assante, M. et al . Enacting open science by d4science. Future Generation Computer Systems 101 , 555–563 (2019).

Coro, G., Candela, L., Pagano, P., Italiano, A. & Liccardo, L. Parallelizing the execution of native data mining algorithms for computational biology. Concurrency and Computation: Practice and Experience 27 , 4630–4644 (2015).

Coro, G., Panichi, G., Scarponi, P. & Pagano, P. Cloud computing in a distributed e-infrastructure using the web processing service standard. Concurrency and Computation: Practice and Experience 29 , e4219 (2017).

Wikidata. SPARQL entity retrieval specifications and examples. https://www.wikidata.org/wiki/Wikidata:SPARQL_query_service/queries/examples (2024).

University of Freiburg. QLever instance. https://qlever.cs.uni-freiburg.de/ (2024).

Bast, H. & Buchhold, B. Qlever: A query engine for efficient sparql + text search. In Proceedings of the 2017 ACM on Conference on Information and Knowledge Management , 647–656 (2017).

Open Geospatial Consortium. Well-known text representation of coordinate reference systems. https://www.ogc.org/standard/wkt-crs/ (2024).

Eurostat. Local Administrative Units. https://ec.europa.eu/eurostat/web/nuts/local-administrative-units (2024).

Brandmueller, T., Schäfer, G., Ekkehard, P., Müller, O. & Angelova-Tosheva, V. Territorial indicators for policy purposes: Nuts regions and beyond. Regional Statistics 7 , 78–89 (2017).

Eurostat. GISCO - the Geographic Information System of the COmmission. https://ec.europa.eu/eurostat/web/gisco (2023).

Eurostat. LAU descriptions in GeoJSON format. https://ec.europa.eu/eurostat/web/gisco/geodata/statistical-units/local-administrative-units (2024).

Eurostat. NUTS descriptions in GeoJSON format. https://ec.europa.eu/eurostat/web/gisco/geodata/statistical-units/territorial-units-statistics (2024).

Commons, W. Wikimedia commons. Retrieved June 2 (2012).

Bartalesi, Valentina. JSON schema of the internal SMBVT data representation. https://dlnarratives.eu/schema%20JSON.json (2022).

W3C Consortium. RDF 1.1: On Semantics of RDF Datasets. https://www.w3.org/TR/rdf11-datasets/#bib-RDF11-MT (2024).

Ciccarese, P. & Peroni, S. The collections ontology: creating and handling collections in owl 2 dl frameworks. Semantic Web 5 , 515–529 (2014).

Meghini, C., Bartalesi, V., Metilli, D., Lenzi, E. & Pratelli, N. Narrative ontology. https://dlnarratives.eu/ontology/ (2024).

Jena, A. Apache jena fuseki. The Apache Software Foundation 18 (2014).

Bartalesi, V. et al . Figshare collection: A Knowledge Graph of European Mountain Territory and Value Chain data. FigShare https://doi.org/10.6084/m9.figshare.c.7098079 (2024).

Openllet. Openllet: An Open Source OWL DL reasoner for Java. https://github.com/Galigator/openllet (2023).

DuCharme, B. Learning SPARQL: querying and updating with SPARQL 1.1 (“O’Reilly Media, Inc.”, 2013).

Lam, A. N., Elvesæter, B. & Martin-Recuerda, F. A performance evaluation of owl 2 dl reasoners using ore 2015 and very large bio ontologies. In In Proceedings of DMKG2023: 1st International Workshop on Data Management for Knowledge Graphs, May 28, 2023, Hersonissos, Greece (2023).

United Nations. 68% of the world population projected to live in urban areas by 2050, says UN. on-line. https://www.un.org/en/desa/68-world-population-projected-live-urban-areas-2050-says-un (2018).

Dax, T. & Copus, A. European rural demographic strategies: Foreshadowing post-lisbon rural development policy? World 3 , 938–956 (2022).

Foundation, T. A. S. Apache jena fuseki. https://jena.apache.org/documentation/fuseki2/ (2024).

The Apache Software Foundation. GeoSPARQL Fuseki. https://jena.apache.org/documentation/geosparql/geosparql-fuseki.html (2024).

QGIS. Software download. https://qgis.org/download/ (2024).

Bartalesi, V. et al . Figshare collection: Visualisation of the MOVING 454 Story Maps. FigShare https://figshare.com/articles/online_resource/Moving_454_Storymaps/25334272?backTo=/collections/_/7098079 (2024).

Bartalesi, V., Lenzi, E. & Pratelli, N. Figshare collection: Fuseki instance for knowledge graph querying. FigShare https://tool.dlnarratives.eu/Moving_454_Storymaps/geosparql.html (2024).

Wikidata. Wikidata Web site. https://www.wikidata.org/wiki/Wikidata:Main_Page (2024).

OpenStreetMap. OpenStretMap data (Protocolbuffer Binary Format) in RDF representation enhanced by GeoSPARQL triples. https://planet.openstreetmap.org/pbf/planet-240701.osm.pbf (2024).

Download references

Acknowledgements

This work has received funding from the European Union’s Horizon 2020 research and innovation programme under the MOVING project (grant agreement no 862739). The authors wish to thank all MOVING-project partners who have contributed to the source data used for the present article.

Author information

Authors and affiliations.

Istituto di Scienza e Tecnologie dell’Informazione “A. Faedo” del Consiglio Nazionale delle Ricerche (ISTI-CNR), Pisa, 56124, Italy

Valentina Bartalesi, Gianpaolo Coro, Emanuele Lenzi, Nicolò Pratelli & Pasquale Pagano

Dipartimento di Scienze Agrarie, Alimentari e Agro-ambientali, dell’Università di Pisa, Pisa, 56124, Italy

Michele Moretti & Gianluca Brunori

You can also search for this author in PubMed   Google Scholar

Contributions

V.B. was one of the main developer of the Narrative Ontology, she orchestrated the experiment and conducted the validation; G.C. designed and developed the workflow for data augmentation and co-orchestrated the experiment; M.M. designed and developed the input MS Excel table collecting all information from the MOVING partners; N.P. designed and developed LAU/NUTS conversion through GISCO; E.L. designed and developed the Story Map visualisation and prepared the OWL graph instance on the Apache Jena Fuseki service; P.P. and G.B. supervised and supported the experiment through the MOVING project funding. All authors reviewed the manuscript.

Corresponding author

Correspondence to Gianpaolo Coro .

Ethics declarations

Competing interests.

The authors declare no competing interests.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary table, rights and permissions.

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/ .

Reprints and permissions

About this article

Cite this article.

Bartalesi, V., Coro, G., Lenzi, E. et al. A Semantic Knowledge Graph of European Mountain Value Chains. Sci Data 11 , 978 (2024). https://doi.org/10.1038/s41597-024-03760-9

Download citation

Received : 22 March 2024

Accepted : 07 August 2024

Published : 07 September 2024

DOI : https://doi.org/10.1038/s41597-024-03760-9

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

Quick links

• Explore articles by subject
• Guide to authors
• Editorial policies

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Semantic Web Services for Multi-Agent Systems Interoperability

• Conference paper
• First Online: 30 August 2019
• Cite this conference paper

• Alda Canito 11 ,
• Gabriel Santos 11 ,
• Juan M. Corchado 12 ,
• Goreti Marreiros 11 &
• Zita Vale 13  

Part of the book series: Lecture Notes in Computer Science ((LNAI,volume 11805))

Included in the following conference series:

• EPIA Conference on Artificial Intelligence

1777 Accesses

6 Citations

Agent-based technologies are often used including existing web services. The outputs of some services are also frequently used as inputs for other services, including other MAS. However, while agent-based technologies can be used to provide services, these are not described using the same semantic web technologies web services use, which makes it difficult to discover, invoke and compose them with web services seamlessly. In this paper, we analyse different agent-based technologies and how these can be described using extensions to OWL-S. Additionally, we propose an architecture that facilitates these services’ usage, where services of any kind can be registered and executed (semi-)automatically.

This is a preview of subscription content, log in via an institution to check access.

Access this chapter

Subscribe and save.

• Get 10 units per month
• Download Article/Chapter or eBook
• 1 Unit = 1 Article or 1 Chapter
• Cancel anytime
• Available as PDF
• Read on any device
• Instant download
• Own it forever
• Available as EPUB and PDF
• Compact, lightweight edition
• Dispatched in 3 to 5 business days
• Free shipping worldwide - see info

Tax calculation will be finalised at checkout

Purchases are for personal use only

Institutional subscriptions

Similar content being viewed by others

Intelligent Agents and Semantic Web Services: Friends or Foes?

Conceptual Integration of Agents with WSDL and RESTful Web Services

Where Are All the Semantic Web Agents: Establishing Links Between Agent and Linked Data Web Through Environment Abstraction

http://www.daml.org/services/owl-s/1.2/generic/ObjectList.owl#List .

http://www.gecad.isep.ipp.pt/epia/19/services/forecast/ .

Lemos, A.L., Daniel, F., Benatallah, B.: Web service composition. ACM Comput. Surv. 48 , 1–41 (2015)

Article   Google Scholar  

Klusch, M., Kapahnke, P., Schulte, S., Lecue, F., Bernstein, A.: Semantic web service search: a brief survey. KI - Künstliche Intelligenz 30 , 139–147 (2016)

McIlraith, S.A., Son, T.C., Zeng, H.: Semantic web services. IEEE Intell. Syst. 16 , 46–53 (2001)

Huhns, M.N.: Agents as web services. IEEE Internet Comput. 6 , 93–95 (2002)

Teixeira, B., Pinto, T., Silva, F., Santos, G., Praça, I., Vale, Z.: Multi-agent decision support tool to enable interoperability among heterogeneous energy systems. Appl. Sci. 8 , 328 (2018)

Pinto, T., et al.: Adaptive portfolio optimization for multiple electricity markets participation. IEEE Trans. Neural Netw. Learn. Syst. 27 , 1720–1733 (2016)

Article   MathSciNet   Google Scholar  

Pinto, T., Vale, Z., Praça, I., Pires, E.J.S., Lopes, F.: Decision support for energy contracts negotiation with game theory and adaptive learning. Energies 8 , 9817–9842 (2015)

Teixeira, B., Silva, F., Pinto, T., Santos, G., Praca, I., Vale, Z.: TOOCC: enabling heterogeneous systems interoperability in the study of energy systems. In: 2017 IEEE Power and Energy Society General Meeting, pp. 1–5. IEEE (2017)

Google Scholar  

Santos, G., et al.: House management system with real and virtual resources: energy efficiency in residential microgrid. In: 2016 Global Information Infrastructure and Networking Symposium (GIIS), pp. 1–6. IEEE (2016)

Santos, G., Pinto, T., Praça, I., Vale, Z.: MASCEM: optimizing the performance of a multi-agent system. Energy 111 , 513–524 (2016)

Kravari, K., Bassiliades, N., Boley, H.: Cross-community interoperation between knowledge-based multi-agent systems: a study on EMERALD and rule responder. Expert Syst. Appl. 39 , 9571–9587 (2012)

Carneiro, J., Alves, P., Marreiros, G., Novais, P.: A multi-agent system framework for dialogue games in the group decision-making context. Presented at the (2019)

Klusch, M., Fries, B., Sycara, K.: Automated semantic web service discovery with OWLS-MX. In: Proceedings of the Fifth International Joint Conference on Autonomous Agents and Multiagent Systems - AAMAS 2006, p. 915. ACM Press, New York (2006)

Lord, P., Alper, P., Wroe, C., Goble, C.: Feta: a light-weight architecture for user oriented semantic service discovery. Presented at the (2005)

Rodriguez-Mier, P., Pedrinaci, C., Lama, M., Mucientes, M.: An integrated semantic web service discovery and composition framework. IEEE Trans. Serv. Comput. 9 , 537–550 (2016)

Greenwood, D., Calisti, M.: Engineering web service - agent integration. In: 2004 IEEE International Conference on Systems, Man and Cybernetics (IEEE Cat. No.04CH37583), pp. 1918–1925. IEEE

Martin, D., Burstein, M., McIlraith, S., Paolucci, M., Sycara, K.: OWL-S and agent-based systems. In: Cavedon, L., Maamar, Z., Martin, D., Benatallah, B. (eds.) Extending Web Services Technologies, pp. 53–77. Springer, New York (2004). https://doi.org/10.1007/0-387-23344-X_3

Chapter   Google Scholar  

Web Service Modeling Language (WSML). https://www.w3.org/Submission/WSML/

Semantic Annotations for WSDL and XML Schema. https://www.w3.org/TR/sawsdl/

Pedrinaci, C., Cardoso, J., Leidig, T.: Linked USDL: a vocabulary for web-scale service trading. Presented at the (2014)

Kopecký, J., Gomadam, K., Vitvar, T.: hRESTS: an HTML microformat for describing RESTful web services. In: 2008 IEEE/WIC/ACM International Conference on Web Intelligence and Intelligent Agent Technology, pp. 619–625. IEEE (2008)

SA-REST: Semantic Annotation of Web Resources. https://www.w3.org/Submission/SA-REST/

Freitas, O., Filho, F., Alice, M., Varella, G.: Semantic web services : a RESTful approach. In: IADIS International Conference WWW/Internet 2009, pp. 169–180 (2009)

OWL-S: Semantic Markup for Web Services. http://www.ai.sri.com/~daml/services/owl-s/1.2/overview/

Christensen, E., Curbera, F., Meredith, G., Weerawarana, S.: Web Service Definition Language (WSDL). https://www.w3.org/TR/2001/NOTE-wsdl-20010315

Hadley, M.: Web Application Description Language. https://www.w3.org/Submission/wadl/

Gunasekera, K., Zaslavsky, A., Krishnaswamy, S., Loke, S.W.: Service oriented context-aware software agents for greater efficiency. Presented at the (2010)

Bellifemine, F.L., Caire, G., Greenwood, D.: Developing Multi-agent Systems with JADE. Wiley, Hoboken (2007)

Book   Google Scholar  

Martin, D.: The Open Agent Architecture

The Zeus Technical Manual. http://zeusagent.sourceforge.net/docs/techmanual/TOC.html

Braubach, L., Pokahr, A., Lamersdorf, W.: Jadex: a BDI-agent system combining middleware and reasoning. In: Unland, R., Calisti, M., Klusch, M. (eds.) Software Agent-Based Applications Platforms and Development Kits, pp. 143–168. Birkhäuser-Verlag, Basel (2005)

Kravari, K., Kontopoulos, E., Bassiliades, N.: EMERALD: a multi-agent system for knowledge-based reasoning interoperability in the semantic web. Presented at the (2010)

Klügl, F., Herrler, R., Klügl, F., Herrler, R., Fehler, M.: SeSAm: implementation of agent-based simulation using visual programming. In: Policy Agents I, II, III View project SeSAm: Implementation of Agent-Based Simulation Using Visual Programming (2006)

Foundation for Intelligent Physical Agents: FIPA ACL Message Structure Specification

Negi, A., Kaur, P.: Examination of sense significance in semantic web services discovery. Presented at the (2019)

Download references

Acknowledgements

The present work has been developed under the PIANISM Project (ANI|P2020 40125) and has received funding from FEDER Funds through NORTE2020 program and from National Funds through Fundação para a Ciência e a Tecnologia (FCT) under the project UID/EEA/00760/2019. Gabriel Santos is supported by national funds through FCT PhD studentship with reference SFRH/BD/118487/2016.

Author information

Authors and affiliations.

Research Group on Intelligent Engineering and Computing for Advanced Innovation and Development, Institute of Engineering - Polytechnic of Porto (ISEP/IPP), Porto, Portugal

Alda Canito, Gabriel Santos & Goreti Marreiros

BISITE - Research Centre, University of Salamanca, Salamanca, Spain

Juan M. Corchado

Institute of Engineering - Polytechnic of Porto (ISEP/IPP), Porto, Portugal

You can also search for this author in PubMed   Google Scholar

Corresponding author

Correspondence to Alda Canito .

Editor information

Editors and affiliations.

INESC-TEC, University of Trás-os-Montes and Alto Douro, Vila Real, Portugal

Paulo Moura Oliveira

University of Minho, Braga, Portugal

Paulo Novais

LIACC/UP, University of Porto, Porto, Portugal

Luís Paulo Reis

Rights and permissions

Reprints and permissions

Copyright information

© 2019 Springer Nature Switzerland AG

About this paper

Cite this paper.

Canito, A., Santos, G., Corchado, J.M., Marreiros, G., Vale, Z. (2019). Semantic Web Services for Multi-Agent Systems Interoperability. In: Moura Oliveira, P., Novais, P., Reis, L. (eds) Progress in Artificial Intelligence. EPIA 2019. Lecture Notes in Computer Science(), vol 11805. Springer, Cham. https://doi.org/10.1007/978-3-030-30244-3_50

Download citation

DOI : https://doi.org/10.1007/978-3-030-30244-3_50

Published : 30 August 2019

Publisher Name : Springer, Cham

Print ISBN : 978-3-030-30243-6

Online ISBN : 978-3-030-30244-3

eBook Packages : Computer Science Computer Science (R0)

Share this paper

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Publish with us

Policies and ethics

• Find a journal
• Track your research

• DOI: 10.69739/jcsp.v1i1.50
• Corpus ID: 271679213

Development and Evaluation of a Web-Based Resident Information Management System

• Uriel M. Melendres , Karlo M. Aranda
• Published in Journal of Computer, Software… 28 June 2024
• Computer Science
• Journal of Computer, Software, and Program

Related Papers

Showing 1 through 3 of 0 Related Papers