ECAI 2004 Conference Paper

[full paper]

Identifying relational concept lexicalisations by using general linguistic knowledge

Maria Teresa Pazienza, Marco Pennacchiotti, Fabio Massimo Zanzotto

The success of the ontological approach in the Semantic Web is strictly related to the possibility of applying in large natural language semantic models. Web documents are first of all documents and the activity of making explicit their content through the ontological language, i.e. extracting concept instances and their relationships, is fairly similar to the classical Information Extraction task. However, to apply Information Extraction models, Semantic Web ontologies need to be equipped with a sort of "linguistic interface" representing the one-to-many mappings between coarse-grained relational concepts of the ontology and the corresponding linguistic realisations. With an eye on the problem of constructing coarse-grained relational concept catalogues, this paper analyses the extent and nature of the general semantic knowledge required for this task analysing how different general-purpose classification algorithms react to the use of this knowledge. For exploiting ambiguous semantic information within the feature vector model, we propose an original model, the semantic fingerprints.

Keywords: Information Extraction, Ontology Learning, Machine Learning

Citation: Maria Teresa Pazienza, Marco Pennacchiotti, Fabio Massimo Zanzotto: Identifying relational concept lexicalisations by using general linguistic knowledge. In R.López de Mántaras and L.Saitta (eds.): ECAI2004, Proceedings of the 16th European Conference on Artificial Intelligence, IOS Press, Amsterdam, 2004, pp.1071-1072.