Synonym Relationship Discovery for Knowledge Organization System Based on Paired Translation Information

摘要

Synonym relationship is one of the most important relationships for knowledge organization systems. In this paper, we propose a series of methods on synonym relationship discovery based on paired translation information already in knowledge organization systems. The translations amount and the translation concurrence strength are inputs of similarity, which then calculated by Boolean comparison, conditional mutual information and cosine similarity of vector space model. With the output of these methods, synonym relationships are discovered and constructed. Precision and richness are used as the indexes to evaluate the methods. With an experiment in new energy vehicles domain, it is found that the effects of the result are acceptable and have the potential to be improved by more translation pairs.