Computing Semantic Similarities Based on Machine-Readable Dictionaries

Abstract

The measurement of semantic similarity is a foundation work in semantic computing. In this paper the authors study the similarity measure between two words. Different from previous works, this paper suggests a novel method that relies on machine-readable dictionaries for measuring similarities. Machine-readable dictionaries are more widely available than other kinds of lexical resources. If two words have similar definitions, they are semantically similar. A definition is represented by a definition vector. Each dimension represents a word in the dictionary. The score of each dimension in the vector is calculated by a variation of tf*idf. Evaluations show that this method achieves competitive results in both Chinese and English.