Research On Community-based Question Retrieval Technology Based On Topic Translation Model

The internet question & answer community has become an important way for people to acquire knowledge and information.It has accumulated a large amount of question and answer data over the years,and how to effectively use these data has become the focus of current research.In related research,the main purpose of question retrieval is to use historical question and answer data to help users find the answers to their concerns,reduce the time for users to wait for others to answer,and avoid the burden of repeated submission of similar problems to the system.This paper focuses on two aspects of question retrieval: question relevance and question & answer quality.It is studied from the following three different points:Firstly,for the problem in the current topic inference based translation model,this paper proposes a model that integrates query topic information.It uses the topic distribution of the query as a disambiguation basis,achieves effective matching of words in queries and candidate questions,and optimizes the original model.Secondly,for the problem that query semantics are not considered when weighting the query terms,this paper proposes a topic-based term weight model.It utilizes the topic model as a semantic mining tool,combines the principle of information entropy,and calculates its weight in the query according to the amount of information included in the word.This model can well solve the verbosity problem caused by complex queries.Finally,for evaluating the quality of question-answer pairs,the paper proposes a scoring model based on user information.It scores the user's authority based on the number of answers the user has taken as the best answer.Then based on the assumption that the quality of the user's published information is positively related to its authority,it uses the authority of the questioner and responder as the quality features of the question & answer pair.The paper uses the learning to rank to combine the quality features and the relevance of questions to form a unified retrieval model.In addition,the experimental results on real data sets show that each model proposed in this paper has achieved excellent results for each problem to be solved.