Domain-Specific Knowledge Acquisition Using WordNet

Abstract

This paper presents a method that acquires new concepts and connections associated with user-selected seed concepts, and adds them to the WordNet linguistic knowledge structure. New domain knowledge can be acquired around some seed concepts that a user considers important. The knowledge we seek to acquire relates to one or more of these concepts, and consists of new concepts not defined in WordNet and new relations that link the concepts with other concepts. The approach consists of forming a corpus with sentences containing seed concepts and then identifying on this corpus lexico-syntactic patterns that reflect semantic relations. The algorithm has four procedures outlined below. Procedure 1: Concept extraction. Input: Noun phrases that contain a seed concept. Output: New concepts constructed around the seed concept. After the sentences in the corpus are parsed, new concepts are sought in the noun phrases where seed concepts reside. This procedure searches in WordNet and other electronic dictionaries for possible concepts in the noun phrases. The final acceptance of the concepts rests with the user. The next step is to create a taxonomy for the newly acquired concepts that is consistent with WordNet. Procedure 2: Classification by subsumption Input: A list of NPs containing the seed as head noun Output: An ontology of concepts under the seed The classification algorithm is based on the simple idea that a compound concept [word, seed] is ontologically subsumed by concept [seed]. Similarly, for a relative classification of any two concepts [word1, seed] and [word2, seed], the ontological relation between word1 and word2, if it exists, is extended to the two concepts. In the case that word1 subsumes word2, then a relation is formed between the two concepts.