Extracting semantic clusters from machine readable dictionaries for an onomasiological search dictionary

The ideal onomasiological dictionary should allow users to search for a word by introducing the knowledge they already have about a concept. Such an onomasiological search relies crucially on indexed information about lexical paradigms or clusters in order to expand the query in the search for the word. This paper describes a semantic clustering method applied to machine-readable dictionaries (MRDs) in order to construct an onomasiological dictionary to find terms from concepts. We first assess related approaches based on ontologies and statistics, before introducing our analogy-based approach that allows us to extract semantic clusters by aligning definitions from two dictionaries. Evaluation of the final set of
clusters for a small set of definitions demonstrates the effectiveness of our approach.