Similar Question And Answer Summary In Community Question Answering

As a new information retrieval and communication platform,question answering community has grown rapidly in recent years.With the development of this platform,many semantic similar questions are proposed repeatedly,resulting in redundant information.At the same time,in multiple answers to a question,the user only selects one of the best answers,ignoring the valid information in other answers.In order to make full use of the questions and answers accumulated in the community question answering,this paper carried out a research work on similar problems and summary of answers,include:1.This paper designs and implements the similarity comparison model based on the multi-view attention mechanism.This paper uses the attention mechanism to learn the interactive information between the text,which is very important for its similarity comparison.In order to improve the expression ability of attention mechanism,this paper designs multi-view attention mechanism to express interactive information from different views.In addition,the model also explores different attention matching functions,similarity comparison functions and model loss functions.2.This paper improves the feature expression ability of the original word vector model.This paper designs a compound word vector to express the word,learning the character information of the word,the semantic information of the word,the part of speech of the word.The character vector is used to learn the word form information,the word part of speech vector is designed to introduce the shallow semantic characteristics,in order to solve the problem of colloquial and unstandardized expression in the community question answering.3.This similarity comparison model based on multi-view attention mechanism is successfully applied in the task of similar question retrieval and answer summary,for question-question and question-answer similarity comparison.In these two tasks,the model of this paper has a certain improvement on the evaluation index.In addition,the compound word vector is used in the task of similar question retrieval,which is better than the original word vector expression.