Subjective Bayes Method for Word Semantic Similarity Measurement

Measuring semantic similarity between words is a classical problem in nature language processing, the result of which can promote many applications such as machine translation, word sense disambiguation, ontology mapping, computational linguistics, etc. This paper combines knowledge-based methods with statistical methods in measuring words similarity, the novel aspect of which is that subjective Bayes method is employed. Firstly, extract evidences based on Word Net, secondly, analyze reasonableness of candidate evidence using scatter plot, thirdly, generate sufficiency measure by statistics and piecewise linear interpolation technique, fourthly, obtain comprehensive posteriori by integrating uncertainty reasoning with conclusion uncertainty synthetic strategy, finally, we quantify word semantic similarity. On data set R&G (65), we conducted experiment through 5-fold cross validation, and the correlation of our experimental results with human judgment is 0.912, with 0.4% improvements over existing best practice, which show that using subjective Bayes method to measure word semantic similarity is reasonable and effective.