Improvements in automatic thesaurus extraction

The use of semantic resources is com-mon in modern NLP systems, but methods to extract lexical semantics have only re-cently begun to perform well enough for practical use. We evaluate existing and new similarity metrics for thesaurus ex-traction, and experiment with the trade-off between extraction performance and efciency. We propose an approximation algorithm, based on canonical attributes and coarse- and ne-grained matching, that reduces the time complexity and ex-ecution time of thesaurus extraction with only a marginal performance penalty.