The Research On Numeric Question Understanding Of English Automatic Question Answering System

With the rapid development of information technology,there are more and more information on the Internet.The search engine has come to be the main tool of information retrieval.However,the information retrieval based on the se-arch engine can't fully reflect the user's query intention,leading to a large nu-mber of irrelevant Web pages to query intention that returns.Besides that,build-ing the query expression is also a burden for ordinary users.Automatic question answering system is an advanced form of information retrieval.It tries to answ-er questions expressed on natural language to facilitate the user's query and im-prove the accuracy of information retrieval.The research purpose of numerical question answering system is to find corresponding numerical answers or corres-ponding calculation formula for numeric problem.This article research is aimed at the hot topic of English numerical question understanding.This paper is based on GORT which is an existing numeric question answ-ering system,aiming at the shortcomings of the link of question understanding,research use machine learning methods to find out the corresponding computatio-n formula for the simple question that inputs into the GORT system.First of all,this paper is to classify questiones.Secondly,doing sentence pretreatment.Final-ly,building "question-formula" sets.For the objective question,through sentence similarity algorithm to find the most simiar question in "question-formula" sets with the objective question.Then the calculation formula of objective question is concluded.This paper presents binary chart and edit distance sentence similarity algori-thmes aimed at operation sentence:The binary chart algorithme consider the ph-rase,The edit distance import replacement semantic function.The experimental results show that the accuracy of question understanding using binary chart reac-hs 80.33%,the edit distance reachs 85.33%.