Numerical Answers Integration Research In Automatic Question Answering System

In today's era of Web information explosion,keyword-based Information retrieval method is not sufficient to fully express the user's query intention,which often results in a lot of irrelevant information being returned.Advanced keyword-based Information retrieval method can improve the accuracy of the query to a certain extent.But the cumbersome process of structure query expression will shut out a large number of ordinary users.After all,ordinary users desire to express the proposed retrieval requirements using natural language.Therefore,automatic question answering system based on natural language naturally becomes a research interest of scholars both in domestic and abroad.Answer integration is one of the key links of question answering system.Its main content makes optimization integration for a set of answers which is returned by the search engine and calculates the ultimate answer.This paper main studies integration algorithm oriented numerical answers.The main job is to relying on an existing numerical question answering system GORT,and improving the defects of answer integration algorithm which is used now.This paper puts forward the concept of predicate unit table and establishes a set of candidate solution processing algorithms based on it.First,the paper define the semantic type of candidate answer set through the predicate unit table,and deletes the candidate answers which semantic type are not consistent with the candidate answer set.Then,calculating unit of candidate answer depends on the semantic type of the query.Last,final answer is calculated according to candidate answer set.The improvement of proposed method is that it can clear the semantics type of candidate answer set and effectively predict the unit of candidate answer.In addition,the paper solves the semantic type inconsistency of candidate answers.We designed and implemented an Automatic Question Answering System of local knowledge base based on the improved Answer Integration Method.Then test andverify the effectiveness of the new method on the basis of Automatic Question Answering System,Finally,we summarizes the research work,and points out the work which also needs to be solved,and points out the research direction and development prospects.