About KAON

KAON was started in August 2001 as an internal research project of the knowledge management groups of the Forschungszentrum Informatik (FZI) and the University of Karlsruhe (TH). Both institutions are lead by Prof. Dr. Rudi Studer.

The aim of KAON was to facilitate the collaboration of both groups and to avoid "reinventing the proverbial wheel". KAON aggregates the implementation deliverables of several governmental research projects and thereby increases the synergy between those projects.

This strategy has proven to be very successful. For example, the quality of KAON implementations could be improved tremedously, since the actual usage of implementations in other projects allowed to fix bugs (otherwise unseen) and contributed additional features.

This lead to near-product quality of the core implementations of KAON, individual KAON modules have been successfully used within several industrial prototypes, which have been developed for different European firms. Some of these clients, e.g. the UN FAO, allowed to contribute code back into KAON, e.g. the support for arabic and chinese character sets.

KAON has also been successfully used to validate particular research ideas that arose in the area of Semantic Web and ontology-based knowledge management, cf. the publications. A particular focus was given to the ontology language used within KAON.

KAON ontology languages

The general goal of KAON is to retain scalability in reasoning with large ontologies and knowledge bases. We therefore followed an incremental and conservative strategy wrt. expressivity of the ontology language supported by KAON.

In the beginning, we supported a simple core language (the so-called "Karlsruhe perspective on ontologies"), which rectified the semantics of RDFS and basically supported the definition of concepts, relations between concepts and partial orders on concepts and relations. Novel, and of central importance for most of our applications, was the support for explicit lexicalizations of concepts and relations. We additionally foresaw the addition of logical axioms to support logical entailment beyond traversal of partial orders. The language was supported by the first version of the KAON API, we developed a simple ontology editor (KAON SOEP), a tool for mapping relational database content (KAON REVERSE) and a prototype for views on ontologies (KAON VIEWS).

We later extended the KAON language (the so-called "Ontology-Instance (OI) models"). to support frequently occurring patterns of logical axioms such as the specification of algebraic characteristics of relations, e.g. symmetry, inverse, transitivity, and constraints on the cardinality of relations. The extension also brought several non-logical features like modularization, strict separation of ontologies and instances and support for spanning objects. The change in ontology language went a long with a complete redesign of the KAON API and the development of new user-level tools, e.g. the KAON Portal and KAON OI-modeler ontology editor.

We currently work on a parallel support for the recently standardized Web ontology language (OWL). We contributed to the development of the OWL API and investigated how we can provide efficient reasoning support for OWL. Our basic strategy is to rely on (deductive) database-techniques to support efficient reasoning with instances.

We will maintain two ontology languages in parallel, since the expressivity of OWL is not needed in many of our applications and it is not needed to break a butterfly on a wheel for those applications. For example, techniques for emerging semantics such as provided by TEXTTOONTO can currently not support the semi-automatic acquisition of complex logical axioms such as expressible in OWL.

Research Themes

Reasoning

The PhD thesis of Raphael Volz shows which fragment of OWL, aka. DLP, can be reasoned with in standard (deductive) databases such as the KAON Datalog engine, XSB and CORAL. We could show that deductive databases outperform tableau-based OWL reasoners by several orders of magnitudes and that the DLP fragment is sufficient to express most (> 80 %) OWL ontologies completely.

The work of Boris Motik investigates how resolution-based approaches (such as employed in deductive databases) can be used to provide reasoning for the OWL Lite and OWL DL variants. He currently develops an OWL reasoner (KAON DL) in context of the EU-funded DIP project.

Text Mining

Semantic Middleware

The work of Daniel Oberle investigates how the underlying concepts of existing Application Servers can be applied and augmented for application development in the Semantic Web. On the other hand, he is trying to incorporate semantic technologies within such a server. The KAON SERVER is the current prototype and distributed in the KAON Extensions project.

Reasoning

Evolution