Research On Questions Similarity Computing In Q&A System

Questions similarity computing is the key in FAQ question answering system. Chinesesentence similarity computing has important relationship with the technology of morphologyanalysis and parsing, however, the current parsing accuracy rate is still not high, and themethod of semantic analysis according to the semantic dictionary is not always precise whilesome unknown words exist. The immature foundation restricted the research on questionssimilarity computing. Chinese questions is a special form of sentences, it has four types.Chinese questions usually have intonation, question words, or the tone of the word and so on.This paper made a thorough research on questions similarity computing by combined thequestions features and morphology analysis technology. In this paper, the work included thefollowing aspects:(1) The paper introduces the existed sentence similarity computing methods at first，thenanalyzing the disadvantages and advantages of each method. At last, Chinese sentencesimilarity computing methods were analyzed.(2) According to the characteristics of questions, combined with morphology analysistechniques, this paper extracted the topic and focus as the key part of a question, then a newalgorithm of questions similarity computing based on Topic-Focus is proposed. This paperdescribes it as3parts including questions preprocessing, questions retrieve and questionssimilarity computing. The innovation point is the method of extracting topic and focus andconsidering its effect in questions similarity computing. The experiment shows this algorithmhas a good result.(3)The proposed algorithm is used in FAQ question answering system to test. This paperdesigns a module of questions retrieval, and introducing the process of retrieval and the keytechnology with detailed description. Then this paper introduces the implementation of eachmodule of the FAQ question answering system including question analyzing, questionsretrieval and questions similarity computing. Finally, a question and answer system within thefield of tourism is come true, the algorithm and system performance well, and the evaluationresults are given.