EIFFEL Ontology Documentation Page

This documentation explores in detail the PIXEL Ontology Tools as one of the main components of the EIFFEL architecture. To access the EIFFEL project website, click here.

EIFFEL aims at revealing the Role of GEOSS as the Default Digital Portal for Building Climate Change Adaptation & Mitigation Applications.

H2020 project EIFFEL is a game changer in the domain of climate change adaptation and mitigation by harvesting the benefits of the GEOSS data. The project will offer to the Earth Observation community the ground-breaking capacity of exploiting existing GEOSS datasets and will also build upon prior knowledge, with minimal new data collection activities. Added-value services interoperable with GEOSS will be designed, using cognitive search and metadata augmentation tools based on Artificial Intelligence, including Natural Language Processing. These tools will leverage advanced cognitive features to extract meaningful information from and enrich GEOSS metadata.

EIFFEL has been financed by the EU Horizon 2020 research and innovation programme under Grant Agreement No 101003518

The term ontology comes from the field of philosophy and relates to the study of being and existence. In computer science, it was first adopted from Artificial Intelligence (AI) researchers in the 1980’s who anticipated its potential for building computational models and allowing some sort of automated reasoning. One decade later, interoperability issues among emerging technologies and platforms introduced an ontology layer as a standard component of knowledge systems.Typically, in computer science the term ontology is defined as an explicit specification of a shared conceptualization. In other words:

• The ontology defines basically the concepts and relationships among them within a particular domain (specific subject or area of knowledge).
• The specification takes the form of a set of classes (representing the concepts) with a set of properties (represented by the properties of each class)

The ontology spectrum describes a range of semantic models of increasing expressiveness and complexity:

• Less structured models: taxonomies (e.g., UNSPSC), database schemas (e.g., SQL schema) and metadata schemes (e.g., WSDL)
• Structured models: thesauri (e.g., WordNet) and conceptual models (e.g. UML)
• High structured models: logical theories (e.g., Ontolingua, semantic web)

Ontologies are normally developed by a team with two main profiles:

• Domain experts: have the knowledge of a specific domain
• Modellers/ontologists: know how to model a new domain, build the semantic properties, create individuals, link with other domains, etc.

The main key benefit provided by ontologies refers to intelligent searches. By means of ontologies, users can organize the information in a way so that software agents (machines) can understand and interpret the meaning; therefore, they might be able to search and integrate data in a much better and powerful way. As the meaning is somehow stored in the ontology, applications can automatically extract data from web pages (e.g., in a semantic web environment), process them and make their own decisions.

Basically, you need more than a conceptual model –typically understood by human beings - if you need machine interpretability (more than machine-processing). Logical theories come in place as strong ontologies, which can be:

• Frame-based: focussed on the entities (like OO modelling), which includes the logical expressions, properties, relations and constraints/rules.
• Axiomatic: not centred on the entities, but rather on the axioms/rules referring to the entities allowing a more distributed approach.

A typical semantic web deployment requires the following modules:

• RDF Parser: it is able to read text in one or more of the previous RDF formats and interpret them as triples. The RDF serializer makes the opposite.
• RDF store: it is a database with the ability to store and retrieve data in the form of triples, also with the ability to merge multiple data sources.
• RDF query engine: it is associated with the RDF store and allows to retrieve RDF data in form of structured queries. SPARQL, standardized by the W3C, is the most common query language.
• Application: it performs some processing with the RDF data, typically obtained from a RDF store via RDF queries.
• Reasoning engine: this module is able to reason and infer logical consequences from RDF data or RDF schemas. Reasoning declarations, or part of them, can be expressed in the RDF query language and therefore the reasoning engine can be part of the RDF database. In other environments, the reasoning engine can also be used as a library by/within the Application.

Before the appearance of NO-SQL databases, W3C provided recommendations, such as Direct Mapping (DM) and Relational to RDF Markup Language (R2RML) to assist in the mapping of a table to RDF and avoid ambiguities or guesses. Currently the concept of RDF store can easily be ported to the web, as many HTML pages include structured information (e.g., contact information, opening hours, ratings, prices, etc.). This capability of embedding structured data in web pages is called rich snippet (vs. plain snippet), and can be really helpful in searches (e.g., to decide for a restaurant among others). Common ways to embed data in HTML web pages are Microdata (Schema.org), RDFa (W3C) and JSON-LD. For example, Facebook uses a simplified version of RDFa to encode OGP (Open Graph Protocol) data in a web page; it allows Facebook users link to pages outside of Facebook.

Besides the way data is embedded (format), it is important to have some shared vocabulary for that data. Schema.org a joint initiative (Google, Microsoft, Yahoo) to provide shared ontologies supporting the embedding of structured data in web pages. It includes common things that you can describe in your web pages, such as Persons, Events, Organizations, Places, Ratings, etc. and binary objects (e.g., audio, video). The namespace for Schema.org is just http://schema.org and one can use the prefix schema: for the different entities. A full analysis of the semantics of Schema.org is beyond the scope of this section, but there is plenty of related literature [6]. Schema.org’s main use is tightly coupled with Search Engine Optimization (SEO) and rich snippets, but any application can make use of it. Typically, the process follows 3 main steps:

1. Define and model your data with Schema.org
2. Embed the data using any format (RDFa, Microdata, JSON-LD)
3. Check it with some tool: Google Structured Data Testing Tool or W3C RDFa distiller

The RDF data model was designed with data federation in mind. The idea of federating information first (producing a federated graph) and then querying the federated information store facilitates the operational logic of the application.

It is not the same data on the web (e.g., a PDF on a web page) that the web of data, composed of linked data. Linked Data are datasets or elements linked across the web; the web of data uses RDF to include data descriptions including links to other data descriptions: - The WEB of documents is built on top pf HTTP/HTTPS, HTML and URLs - The WEB of data is built on top of HTTP/HTTPS, RDF and URIs/IRIs

Currently resources (URLs/URIs) on the web can be pretty much anything:

• URI for Valencia in DBpedia: https://dbpedia.org/page/Valencia
• URI for Valencia in Wikidata: https://www.wikidata.org/wiki/Q8818
• URI for the name Joaquin Sorolla in the Library of Congress: https://www.loc.gov/item/2016857349/
• URI for the “1st of January 2021 at 17:00 -local time- in New York”: https://www.timeanddate.com/worldclock/meetingdetails.html?year=2021&month=1&day=1&hour=17&min=0&sec=0&p1=179
• URI for a publication using ISO DOI: https://doi.org/10.1109/ACCESS.2018.2798918

Depending on the covered domain or topic, there are literally hundreds of possible schemas published on the Web of data. To find schemas there are Web applications providing directories and search engine functionality, such as Linked Open Vocabularies. Following the success of REST (Representational State Transfer) for web services, the W3C defined the Linked Data Platform (LDP) in 2015 to specify how Web applications can publish, edit, and delete data resources using the HTTP protocol. LDP distinguishes two kinds of resources (RDF and non-RDF) and introduces the notion of container (basic, direct, indirect) to manage resources.

OWL is an ontology language for the Semantic Web with formally defined meaning; it provides classes, properties, individuals, and data values, which are typically stored as Semantic Web documents. The OWL recommendation is currently on its version 2.0, extending OWL v1 but preserving, at the same time, the basic principles. Whereas RDFS-Plus is a subset of OWL, OWL itself is more powerful and allows specifying many more aspects while relating entities, e.g., by defining restriction classes. In fact, OWL supports various types of restrictions, such owl:allValuesFrom, owl:someValuesFrom, owl:hasValue. Cascading restrictions and RDFS can be used to model complex relationships between classes, properties and individuals. OWL also extends RDFS-Plus with a full set of theory language, including intersections, unions, and complements (owl:instersectionOf, ow:unionOf, owl:complementOf, owl:oneOf, owl:cardinality) . They can be used in various ways:

• Combine restrictions (e.g., set of CC applications that work with at least one ECV), including cardinality
• Define restrictions (e.g., a Dummy application is a CC application that has no real data)

Due to governance and flexibility, ontologies can be modularized. OWL provides specific built-in classes to help in this way (e.g., owl: Ontology, owl:imports) with no inference purpose in mind. An example about this is the QUDT (Quantities, Units, Dimensions and Types) ontology which, unlike OGP and Schema.org, is able to cut across domains allowing, for example, unit check and unit comparisons, applying unit conversion if possible.

OWL 2 also includes other subsets or profiles of the OWL language identified for various practical technological reasons, often having to do with how OWL relates to other technology. The profiles are called OWL 2 EL (restricted to improve computational complexity), OWL 2 QL (restricted to be compatible with database queries), and OWL 2 RL (restricted to be compatible with Rules processors).