Reasearch And Implementation On Question Retrieval And Answer Extraction Of Online Community Question Answering

Dued to the complexity of natural language,the existing search engine could not deal with user's natural language questions effectively.Compared with the search engine,Community-based Question and Answering(CQA)system can answer natural language questions directly with high quality answers.Community-based Question and Answering system has become an indispensable and important source of high quality information.However,there are the following challenges of Community-based Question and Answering system.Question answering resources scattered in various Q&A platform;various expression of natural language questions caused similar questions difficult to match;high quality answers submerged in a large number of low quality answers;best answer of similar questions cannot quickly achieve.In this paper,we try to solve the problems in question retrieval and answer extraction.We design and implement a Q & A system with multi platform of online Q & A community.In the part of the question retrieval,we designs and implements the automatic identification of problem types and classification to the specific online question and answer community.Firstly,the system deal with the questions of natural language form and extract the features of the questions,consider the corresponding relationship between different question types and different online Q&A communities.The similarity questions of multiple sources is then sorted according to the correlation degree.Secondly,Considering the questions of semantic distance feature,statistical model characteristics and quality characteristics of questions,sort questions through learning algorithm to adjust the weights,to achieve a multi platform similar questions scheduling system.In the part of the answer extraction,we proposes the method of combining rule pattern recognition and supervised machine learning to filter the spam in the answer.At the same time,considering the characteristics between questions and answers which could recognize the answer quality,the list Net algorithm is used to build an answer ranking model,to select the most relevant answer to the question.The method is evaluated on the data set and the artificial test set on Yahoo Answers!,and compared with the classical algorithm.Experimental results show that the system can achieve multi platform online question retrieval and answer extraction,which basically meet the needs of users.Based on the above two key technologies,we participated in the design and development of a QA system.The system participated in an online Q&A contest organized by the 2015 international text retrieval conference(TREC),and achieve 3rd score in the answer quality part.