A Synonym Extraction Method Based on Intimacy*

Abstract

This paper proposes a new concept—intimacy for synonyms extraction. Firstly, we find the first n similar word labeled A1...An of the word A, and then, find the first n similar word groups of A1,A2...An respectively, and calculate the intimacy based on whether or not A appears in these word groups and the position A appears to obtain the final intimacy score. Finally, sort the final intimacy score and obtain the first n synonyms of A. In this paper, we test on Wikipedia dataset, and compare with word similarity computing tasks. Experiments show that the synonyms found by this method are higher than other methods both in subjective and objective evaluation.