Design And Implementation Of Community Questions Retrieval System

With the arrival of the Web2.0, Community Question Answering(CQA) as a full of interactive information access method, and gradually become an important platform for the user to obtain information and experience. As the user centered method of the access new question and answer, because of its answer is usually through artificial screening, with higher quality, compared to the traditional question answering system is more accurate, but because the answer is not real-time, can not provide users with the best user experience, then lead to the community develop slowly in these years. Then more and more researchers focus on how to search for the existing problems, find out the problem and return to the user, in order to achieve the purpose of real-time, that is, Community Retrieval Question, Community Question Answering Retrieval System is based on this.Lots of Community Question Answering Retrieval System is running, but existing Community question answering retrieval system mainly includes two parts of the problem: 1) due to the current system’s synonym extension modules are all rely on dictionary, this may cause expansion and semantic inconsistent problems; 2) because the user input is described in natural language sentences and their complex structure and long sentence may cause difficulties in finding important word in the sentence. This paper mainly according to these drawbacks, design and implementation of a new Community Question Answering Retrieval System, use word2 vec expansion term in order to measure the similarity to improve the synonym expand without semantic problem, also using the dependency relation is found important word in the sentence, thereby enhancing the community question and answer retrieval system. The main research content of this paper is as follows:Firstly, according to the existing knowledge, the system of community question answering retrieval system is constructed. The architecture, function module and system detail process are designed, and the algorithm flow and implementation process of each functional module are introduced. The most important part of the retrieval system is focus on solving the existing problems.Secondly, this paper proposes a new method for the extension of word2 vec, according to question existing term expansion methods that expansion words with the original words with the same and may cause by inconsistent semantic and fusion word2 vec the word expanding method, this paper proposes a new method based on the traditional synonym expansion,and use word2 vec introducing semantic extension then calculate the similarity, and the two ways to get the expansion words and similarity fusion obtained new word item extension set and its similarity set, in order to use it to calculate the extended term of term weighting, which is used to improve the effectiveness of the retrieval system.According to the current community question answering question retrieval system in the retrieval model does not consider the question of complex structure and long sentences may cause finding the question important lexical entry difficulties, in this paper puts forward the question dependency relation retrieval method based on important degree, innovation is different according to the questions set up an important measure of the relation to association between questions in a tight, then an important measure applied to the weighting mechanism in the lexical entry, from lexical entry weight to find important lexical entry questions, then according to the extended term weighting fusion word2 vec expansion of the word and similarity, the searching model and extends the results in combination with again in descending order, get the final retrieval results to improve retrieval system.Finally, experiments are designed to verify the effectiveness of the proposed method, and then give the system user interface to demonstrate the availability of the system.