Meronymy Extraction Using An Automated Theorem Prover

In this paper we present a truly semantic-oriented approach for meronymy relation extraction. It directly operates, instead of syntactic trees or surface representations, on semantic networks (SNs). These SNs are derived from texts (in our case, the German Wikipedia) by a deep linguistic syntactico-se mantic analysis. The extraction of meronym/holonym pairs is carried out by using, among other components, an automated theorem prover, whose work is based on a set of logical axioms. The corresponding algorithm is combined with a shallow approach enriched with semantic information. Through the employment of logical methods, the recall and precision of the semantic patterns pertinent to the extracted relations can be increased considerably.