Improving the Syntax-based Retrieval System Using Collocation Indexing

Abstract

The purpose of this paper is to design a syntax search system and to apply it to a movie search system. The concepts applied include those in the field of linguistics and collocation, to increase the speed of the syntax search system. First, we must process the keywords in the database by labeling them according to their part of speech. From the results of the process, we will construct a K-gram index and Collocation index.In this proposal we bring out a few examples of common English syntax rules and sentence structures as test models. After the run through, the K-gram index and the Collocation index are compared. We have found that part of the sentence, after having gone through the Collocation index search, has a far smaller sample space that the K-gram index alone, which is to say that the Collocation index is able to find the most correct result from fewer samples, thus minimizing the time cost in Query Match.