Research Focus

The Semantic Web and Ontological Engineering (SWOE) group at KIZI (one of its four research groups, over-arched by the virtual Knowledge Engineering Group) undertakes research in knowledge and data representation and management, in particularly in connection with the semantic web. The focal areas of the group currently are:

  • Linked data modeling, publishing, matchmaking and exploitation, especially for public sector, e-commerce and encyclopedic (Wikipedia etc.) domains
  • Ontology and data vocabulary design and management
  • Ontology and data vocabulary analysis, evaluation and matching
  • Ontology and linked data visualization
  • Linked data mining

Notably, we use the term ‘ontological engineering’ as short for ‘ontology engineering and ontological data engineering’, i.e., we are not only interested in the proper design of ontologies (vocabularies) but also in their adequate use in describing data of various kinds.

Pre-history

The seed of the group had been gradually formed since approx. 2000, the initial topics having been the computerization of clinical guidelines (e.g., in the EU FP4 MGT project, where the Stepper tool for text-centric guideline formalization was developed) and modest contributions to the initial phase of the semantic web research (via non-funded partnership in EU FP5 projects such as OntoWeb and Knowledge Web). In EU FP6, the team had been involved, among other, in the design of the Core Ontology of Multimedia (COMM), in the frame of the K-Space Network of Excellence (2006-2008). The group also hosted the EKAW 2006 conference (with V. Svátek as Co-Chair).

Present

  • An ongoing activity, started around 2006, is ontology tool benchmarking, with special focus on Ontology Matching. Most notably, O. Zamazal has been engaged in this community as co-organizer of the OAEI campaigns, as main provider of the ‘conference’ benchmark (OntoFarm dataset) used in these campaigns, as well as the author and maintainer of the OOSP tool allowing for building multi-ontology corpora (as potential benchmark collections) on the fly.
  • Since 2010, the SWOE flagship in software artifact development has been the PatOMat project, which provides a rich machinery for pattern-based transformation of OWL/RDFS ontologies and data vocabularies. PatOMat components are in use by several external parties, and embedded into tools such as ORE (Univ. Leipzig) or XDtools (ISTC/CNR Rome). PatOMat use cases include unification of ontology/vocabulary modeling style (for example, for e-commerce ontologies), ontology complexity downgrading, ontology adaptation to Linked Data best practices, or support of matching or merging of ontologies with heterogeneous structure. User-oriented tools, such as the GUIPOT Protégé plugin for OWL2OWL pattern-based ontology transformation, have been developed.
  • A closely interconnected effort is that of applying ontological background models (OBMs), expressed in a modeling language called PURO, in ontology analysis and design. The central artifact thereof is PURO Modeler, a graphical editor of OBMs allowing to build ontology fragments in various OWL styles through the associated OBOWLMorph tool. This view of OWL ontologies, which abstracts from the particular encoding style, also led to new research sub-project on studying the focused categorization power of ontologies across multiple OWL encoding patterns.
  • In recent years, linked data have become another major topic for the group. In 2011-2014 the group had been responsible for a particular use case in the EU FP7 LOD2 project, namely, that on Public Procurement. A Public Contracts Ontology has been designed and numerous RDFization efforts took place (mostly in collaboration with the XRG group at Charles University, Prague, using LOD2 Stack tools such as UnifiedViews). A follow-up EU project was H2020 OpenBudgets.eu, devoted to RDFization (mostly via LinkedPipesETL) and subsequent analytics of public budget and spending data.
  • A related effort in linked data processing deals with linked data analytics, using both data mining tools and visualization tools (such as our LODSight). In 2013, the first edition of the Linked Data Mining Challenge was organized, with significant involvement of the SWOE team, and then another three, associated to the Know@LOD workshop (2014, 2015 and 2016) collocated with the ESWC conference. Under design is currently a tool for mining rules from RDF graphs, called RDFRules.
  • SWOE is also managing the Czech DBpedia and develops applications (lately, in particular, games such as DB-quiz by J. Mynarz) that exploit both the Czech and English dataset.
  • Finally, the education of linked data is now being supported by SPARQLab, a web-based exercise-book for SPARQL, already in use for the 4IZ440 – Linked data on the web course, and awarded as Best Student Application in the Czech OpenData Application contest.

As summary, the current core vision of the group is:

  • To promote the linked data and open data principles (with special respect to the public sector), and train specialists for dealing with such data
  • To help bridge the gap between the linked data and ontological world using a “reactive ontological engineering” approach: trying to make sense of data structures that people create intuitively rather than enforcing them a “single best model”
  • To investigate the interplay between different modalities of web data/schema semantics: natural language, graph structure, logical entailment etc.