The OntoBasis Project
The 3rd Ontobasis Workshop: CNTS - Universiteit Antwerpen
Automatic initiation of an ontology: how far can we go?
Today, a continual increase of the amount of available electronic resources, associated to more and more powerful computers and to the development of techniques and tools in natural language processing open broader views for text processing. More specifically, the improvement of information extraction techniques, in terms of efficiency and reliability, allows to focus on unsupervised processes for information retrieval tasks, including the automatic building of repositories of semantic knowledge for the construction of domain specific ontologies.
In the context of the OntoBasis project, we have worked on different language domains and extraction techniques, focusing on improving the quality of the syntactic analysis as well as the accuracy of the semantic information extraction step that follows. We will present a summary of the sets of experiments we have carried out as well as a comparison of the various results, in regard of the quality of the data and the size of the corpora provided for each domain. We will show that an automatic bootstrapping of a domain specific ontology is possible, and that there is still space for improvement at each level of the extraction process.
The OntoBasis Project. © 2002 - 2004