Towards cultural knowledge evolution: experiments with alignments repair
报告人 : Pr Jérome Euzenat (LIG,Inria)Knowledge, as expressed formally in the ontologies used on the semantic web, is pressured to evolve. This evolution may be approached through detecting and propagating changes. It can also be considered under a more opportunistic strategy of repairing it when it breaks. Cultural evolution refers to the application of the theory of evolution to culture. In informatics, it has been successfully applied to agents evolving natural language and we study its relevance to knowledge evolution. We will report on experiments developed for testing this approach in the domain of ontology alignment repair. Alignments between ontologies may be established through agents holding such ontologies attempting at communicating and taking appropriate actions when communication fails. We show how agents, using very simple adaptation operators for locally modifying their alignments when they detect incorrect correspondences, are able to reach remarkable global alignment quality. We will discuss new interpretation of these results and future work on the topic.
cedric.herpson (at) nulllip6.fr