Research And Implement Of Bilingual Question Answering System Based On Structured Data

With the rapid development of the information age and web2.0 technology, structured data has been greatly enriched. Because of the inherent shortcomings of traditional search engines, question answering system has become an increasingly hot research direction. The structured data has a high ease of use, higher reliability. Therefore, the study of more efficient and practical question answering systems based on structured data, has important significance and practical value.This paper aims to key technologies involved in question answering system based on structured data, and to implement a bilingual automatic answering system based DBpedia triples and Baidu Encyclopedia data. The main contents and innovations are as follows:1) A detailed analysis of the types and status of the question answering system. Analyze the key technologies involved in QA system and describe the research background and significance;2) Propose a question analysis algorithm based on dependency tree to extract relation phrases and mentions from questions, and Propose several heuristic rules to improve the extraction performance. Use machine learning methods to improve the performance of coreference resolution. Propose and implement candidate nodes recall and query expansion algorithm using suffix tree and a rule-based filtering algorithm. Implement an entity link learning algorithm using learning to rank technique;3) Propose an answer extraction algorithm by combining the sub-graph matching and SPARQL statements. Proposed semantic properties based matching rules to improve sub-graph matching algorithm. For general questions, we deploy sub-graph matching technology to ensure system efficiency; and for comparative or superlative questions, we find the answers by generating SPARQL query.4) In order to solve the problem of lack of Chinese structured data, we introduce English Knowledge to help answer Chinese questions by a translation module. We design and implement a bilingual question answering system based on structured data, and verify the system performance through experiments.