Andreu Sulé, sule@ub.edu
Lecturer in the Department
of Library and Information Science and Audiovisual Communication at the University of Barcelona
Laia Lapeyra, laialapeyra@gmail.com
Student in the Bachelor’s Degree in Information and Documentation at the University of Barcelona
Andreu Sulé is the coordinator of a line of research on the implementation of Semantic Web technologies (RDF, RDF Schema, SKOS and OWL) in digital collections in libraries, archives and museums in Catalonia and the rest of Spain. His personal web page is http://bd.ub.edu/pub/sule/. Lecturer at the Faculty of Library and Information Science of the University of Barcelona from 1998, his subject of interest are Organization of Information, Metadata, Encoding Standards, Vocabulary Control, Semantic Web and Design of Retrieval Systems.
Laila Lapeyra is a student at the Faculty of Library and Information Science of the University of Barcelona. Currently studying the third course of the library and information science grade. Since she is just starting in her professional life there's a lot of things She is interested in, from preservation of documents in archives, to public libraries, but the subject she is most interested in are the Semantic web and all the elements envolved in it.
This paper is an introduction to the Semantic Web and Linked Data, their theoretical foundations, the technologies used to develop them, and the main applications that exist today. The paper concludes with a set of success stories that exemplify the use of the Semantic Web and Linked Data in the real world.
1. What is the Semantic Web ?
When one thinks of semantics, what comes to mind is the area of linguistics that studies the meanings of words. Extrapolating this idea to the Web offers an initial definition of the Semantic Web as an extension of the Web in which data, in addition to having meaning for humans, also have meaning for machines in such a way that they can be searched, found, interpreted, shared and reused among applications, organizations and communities. This is known as the Web of data.
An example can be found in the Berners-Lee, Hendler and Lassila’s foundational article, which presents a case in which automated agents can find the specialist physician closest to the home of Lucy and her brother’s mother. Thanks to the Web of data, these agents also have access to the physician’s schedule and the mother’s schedule and can even arrange the date and time of a medical appointment, also without human intervention. All of this can be achieved in a matter of minutes thanks to the Semantic Web.
2. What is Linked Data?
Another important concept within the Semantic Web is Linked Data. The aim of Linked Data is to relate data described using the RDF model (Resource Description Framework) so that machines can browse the Web.
An example of Linked Data is DBpedia, a project to convert all Wikipedia content into RDF and link this content to other databases like GeoNames. By incorporating other data, the data offered by DBpedia become more complete and precise.
The figure below shows DBpedia and other datasets and all the links that exist between them:
Figure 1. Map with linked RDF datasets
In 2006 Tim Berners-Lee, inventor of the Web and initiator of Linked Data, defined four principles for how linked data should be published :
1. Use Uniform Resource Identifiers (URIs) as names for things.
2. Use HTTP[1] URIs so that people can look up those names on the Web.
3. When someone looks up a URI, provide useful information, using the standards (RDF, SPARQL [see in Section 3 below]).
4. Include links to other URIs, so that they can discover more things.
At present, one might think there are numerous linked data when, but there really are documents linked (hypertext). For example, this link could make one think that it is Linked Data because it provides access to a definition of Linked Data. Actually, though, what is linked is a document, not the data that it contains.
Linked data that are available to the public and can be reused and republished without any restrictions are referred to as Linked Open Data (LOD). Tim Berners-Lee suggested a five-star deployment scheme to evaluate LOD :
Available on the Web (whatever format) under an open license[2].
Published as structured data (e.g., Excel instead of an image scan of a table).
Use of a non-proprietary[3] open format (e.g., CSV instead of Excel).
Use of URIs to denote things, so that people can point at them.
Linking of your data to other data to provide context.
3. Technologies of the Semantic Web
In our definition of the Semantic Web, we said that Web data must be understandable to other machines, just as they are to people. It is necessary, therefore, to describe the data so that machines can understand. With this aim, the W3C (World Wide Web Consortium) created the RDF standard, which provide a framework to describe and publish resources on the Web.
The basics of the RDF are relatively simple. Resource descriptions are expressed in triples made up of Subject, Predicate and Object. For example, the representation of “Marcel Proust is the author of the work À la recherche du temps perdu” would be:
- Subject: À la recherche du temps perdu.
- Predicate: creator.
- Object: Marcel Proust.
The RDF model gives priority to identifying the different components of a triple with URIs. In this way, the triple in the example above would become:
- Subject: http://fr.dbpedia.org/page/%C3%80_la_recherche_du_temps_perdu
- Predicate: http://purl.org/dc/term/creator
- Object: http://fr.dbpedia.org/page/Marcel_Proust
Using URIs instead of literal values ensures that resources are identified unequivocally, avoiding any of the problems of synonyms, polysemy, different idiomatic forms, etc. that are typical of natural language. In this way, machines can read the semantic value of resources unambiguously and establish relationships among the same resources described in different places or datasets.Figure 3. Triple with URIs for the sentence “Marcel Proust is the author of the work À la recherche du temps perdu”
For this theoretical model to be understandable by machines, however, the triples must be encoded (i.e., serialized) with a language that is machine-readable. RDF can be encoded with formats like Turtle, JSON or N-Triples, although one of the most common is XML (Extensible Markup Language). Because of its high degree of standardization (it is a W3C recommendation), the use of XML facilitates the exchange of RDF-structured information among different computers with different operating systems. Serialized using XML, the triple in the example would become :
<?xml version=”1.0″ encoding=”utf-8″?>
<rdf:RDF xmlns:dc=”http://purl.org/dc/elements/1.1/” xmlns:rdf=”http://www.w3.org/1999/02/22-rdf-syntax-ns#”> |
Another technology of the Semantic Web is SPARQL (SPARQL Protocol and RDF Query Language), an RDF query language that enables to look up Semantic Web resources using different data sources.
Lastly, mention should be made of OWL (Web Ontology Language), which is another W3C language used to define ontologies, that is, to explicitly represent the meaning of things and groups of things, as well as their relationships. In this way, machines can more precisely understand the meaning of resources and how they are related. They can even make non-explicit inferences in the descriptions. For example, if the domain is “Animals”, the ontology would represent the meaning of each animal and its related animals (by species, by family, etc.).
4. Published semantic data
Even though the automated agents conceived in Berners-Lee, Hendler and Lassila, (2001) do not yet exist and the people in their example would actually have to make a number of phone calls to locate the best medical specialist and then arrange an appointment, it is worth mentioning that there have been significant advances in recent years in terms of the amount of semantic data available on the Web and the number of applications and services based on semantic technologies (RDF, RDFa, RDF Schema, OWL, SPARQL, etc.).
The Web has lately seen an increase in the number of semantic data, be it through the original encoding of Web content (RDFa) or through the transformation of data stored in other standards (e.g., relational databases). As for the use of RDFa, there were some 200 million Web pages with RDFa data as of November 2015 according to information from the Web Data Commons portal. This figure is 36.26% of the total number of Web pages containing structured data. Indeed, two of the vocabularies most commonly used today to describe Web page content, Schema.org (Google) and Open Graph (Facebook), can be encoded with RDFa.
It is also highly significant that there has been an increase in recent years in the number of datasets marked up semantically. In the library community, for example, several organizations have created mechanisms to offer their data in RDF format, typically through the transformation of their original model (MARC21, XML-EAD, Dublin Core, etc.) into a semantic format and the creation of SPARQL endpoints. At present, a variety of major world libraries have joined this initiative, including Europeana and the Library of Congress and institutions in Spain, Sweden, and also France, the UK and Germany. Specifically, the National Library of France created the portal data.bnf.fr in 2011 to group together data from various sources (its general catalogue, archives and manuscripts, its digital library Gallica, etc.) as well as their links to external resources (Le thésaurus W, VIAF, the Sudoc catalogue etc.). The portal data.bnf.fr offers:
- permanent URIs for resources;
- presentation in RDF of the data available on each Web page and of the database as a whole (in various syntaxes: RDF-XML, RDF-N3 and RDF- NT);
- a SPARQL service for data queries.
Figure 4. Functional diagram of data.bnf.fr
Another sector where the semantic model is gaining prominence is in the open data provided by governments. The US, UK, Sweden, Spain, France and the European Data Portal are all offering open metadata on public sector information using the RDF model and others. For example, the Spanish government’s portal contains 1,716 RDF datasets expressed in different serialization formats (Turtle, XML, N3).
Figure 5. RDF-XML file of the address and telephone directory for Spain’s Ministry of Justice
In the area of Linked Data, the data offered by the Mannheim Linked Data Catalog are available. According to this catalogue, there were 1,920 datasets published as linked data in 2014 and they range from major repositories like DBpedia, Lexvo and Freebase to specific services like Calames, the collective catalogue of French archives and manuscripts.
Figure 6. Home page of the Mannheim Linked Data Catalog
5. Applications and success stories
In terms of applications and services based on semantic technologies (RDF, RDFa, RDFS, OWL, SPARQL, etc.), the W3C consortium maintains two portals, Semantic Web Development Tools and Semantic Web Case Studies and Use Cases, which bring together development tools for Semantic Web applications and descriptions of successful cases.
The tools are numerous and highly varied. They include utility tools to store and search RDF triples (e.g., OpenLink Virtuoso), ontology editors (e.g., Protégé), ontology and semantic document browsers (e.g., Swoogle), generators of semantic descriptions (e.g., FOAF-a-Matic) and visual displays of resources expressed in RDF (e.g., LODLive).
Figure 7. Graphic representation of the concept “Paris” in DBpedia using LODLive
Semantic Web Case Studies and Use Cases provides information on 35 services in which semantic technologies have been applied successfully. Examples range from classic solutions, such as the inclusion of a semantic browser on an organization’s Web page (e.g., Zaragoza City Council in Spain and Electricité de France), to much more innovative cases, such as the use of the Semantic Web in contemporary dance studies (Liverpool John Moores University). Especially significant for the importance of this resource is the case of BBC Music, which by using the same keys employed by MusicBrainz to identify artists can aggregate data from that open music encyclopedia in its own Web content.
Figure 8. BBC Music’s Web page on Prince
6. Conclusions
The aim of the Semantic Web and Linked Data is to take the Web to the next phase: the Web of data, where data are understood not only by humans, but also by machines. Thanks to this understanding, machines will be able to make an extremely important qualitative leap in the way they exploit Web data.
The technologies to make this new phase possible already exist (RDF, OWL, SPARQL, etc.) and they are backed by an internationally renowned organization: the W3C. Now, the need is to increase the amount of semantic data on the Web and develop the tools needed so that the Semantic Web and Linked Data push beyond the current success stories, which are largely limited to academic experiences and very specific communities, and make the leap to the general public with services that improve people’s daily lives.
Bibliographical notes
[1] Hypertext Transfer Protocol: Protocol for transferring hypertext and multimedia documents on the Web. For more information, see point 1.1 at https://tools.ietf.org/html/rfc2616#section-1.1
[2] An open license is one that permits software modification and redistribution. For more information: http://opendefinition.org/guide/
[3] Non-proprietary formats can be used for whatever purpose you want and modified as you wish. Furthermore, no payment is required for their use. For more information: http://www.linfo.org/free_file_format.html
Une réflexion sur « Introduction to the Semantic Web and Linked Data »