Research On Answer Retrieval Method Based On Question Semantic Extension

Automatic question answering system is one of the most important research directions in natural language processing.It can obtain timely and effective information from massive data.The automatic question answering system not only shortens the query process,but also improves the efficiency.This paper will focus on the information retrieval module and answer extraction module,combined with the research status at home and abroad.And it is research to in the aspects of feature extraction,question semantic extension,answer extraction and so on.In the process of using questions to retrieve answers,there are two main problems: one is that the information retrieved is not comprehensive due to the short question sentences;the other is that the scale of document candidate set returned by the information retrieval module is too large,which results in the decrease of the accuracy of answer extraction.To solve the first problem,we need to extend the semantics of questions.This paper uses the technology of automatic encoder to solve this problem.First of all,we need to preprocess the question text,including segmentation and removal of stop words;then extract the key words of the processed question word set,and quantify the words;At last,the key words of the questions are trained by the technology of automatic encoder to prepare for the later answer extraction.To solve the second problem,based on Lucene inverted index,this paper initially selects the set of candidate answer sentences that may contain answers in a large number of candidate answer documents,and then calculates the similarity between the question sentence and the candidate answer sentence to determine the final answer.The evaluation method of the experiment adopts the comparison method,that is,in the case of the same classification standard and retrieval method of the question analysis module,one group of answer retrieval experiments without semantic extension is compared with another group of language extension experiments realized through the content of this paper,which proves the feasibility and effectiveness of the experimental content of this paper.