Short Text Similarity Computing Method towards Agriculture Question and Answering Systems

摘要

Text similarity computing is the core issue that question-answering system needs to solve. It is mainly used to filter out the existed problems which are similar to the user's questions from database. Because of the low recall of domain keywords in domain text similarity computing based on traditional semantic dictionary, this paper proposed a short text similarity computing method in the field of agriculture based on the extended version of Tongyicicilin which referred to as CiLin. This paper propose to consider both the similarity and correlation when calculate the words' final similarity. The experimental results show that the proposed short text similarity computing method resolve the problem of the low recall of domain words in traditional semantic dictionary well, and improve the similarity calculation performance of high relevant keywords greatly.