Chinese analogy search considering multi relations

Abstract

There are some specific relations between the entities of the nature. The same relationship exists between two pairs of entities, simply to say, what A is to B as C is to D. For two entities (A,B), we cannot say precisely the relations between them, and given an instance C similar to A, we cannot figure out the entity D which corresponding to B. Analogical search provides a new way of searching D after given the query (A,B) and (C,?). This paper proposed a Latent relation search method based on words frequency and weight of words. We first explore all the relations between two entities A and B using K-means clustering method, then we determine D corresponding to C in various relations. The proposed method achieves an MRR of 0.773. © 2012 IEEE.