Research And Application Of Question Classification And Answer Evaluation In Community Question Answering System

In recent years,with the development of the Internet,traditional Q&A platforms have become increasingly unable to meet people’s needs due to content abundance and Q&A quality.As a result,community Q&A sites between encyclopedia and traditional Q&A have emerged.Most of the community Q&A websites do not rely on authoritative organizations to provide answer information,but use the way community Q&A users generate content to provide information,and users play the role of information production,reception and dissemination.In view of the characteristics of community Q&A,the quality of website content may also be affected by the quality of users.The main purpose of the question analysis of the community question and answer system is to clarify the user’s intention to ask questions,which is the premise for the effective positioning of the correct answer.Question classification is an important part of the question analysis;answer quality evaluation is related to the community question and answer system,which ensures the high-quality content in the system can be spread and pushed.Therefore,the research on the classification of questions and the evaluation of answer quality in the community question and answer is of great significance to the improvement of the quality of the community question and answer platform and the promotion of question recommendations.Specifically,this thesis mainly completed the following tasks:1.Aiming at the characteristics of questions in community question and answer,a model based on BERT-IndRNN-Attention is proposed.By inputting the question text into the BERT fine-tuning model to obtain the word vector,the word vector with sufficient semantic information can be obtained,and then put the word vector in IndRNN(Independent Recurrent Neural Network,IndRNN)to solve the problem of "gradient disappearance" and "gradient explosion" in the propagation process of RNN(Recurrent Neural Network,RNN),and use the Attention mechanism to enhance the contribution of keywords to question classification.Finally,it is classified in the softmax layer.This method effectively improves the accuracy of question classification in community question and answer.The experimental results on Baidu dataset show that the classification effect of BERT-IndRNN-Attention model is significantly improved compared with other network models.2.The existing evaluation methods of answer quality rely more on feature engineering for experimental research,but feature engineering requires a lot of effort to formulate features,and there is a problem of insufficient semantic information of the question and answer text.This thesis regards answer quality evaluation as an answer quality classification problem,this thesis propose an answer quality classification method based on BERT-CNN.The word vector is obtained by inputting the question and answer pair into the BERT model,and then the word vector is input into the CNN model to obtain semantic features.Finally,the answer quality is classified in the softmax layer.This method effectively improves the effect of answer quality prediction.This thesis uses the Sem Eval2015 Task3 data set to conduct experiments,and the results show that the classification effect of the BERT-CNN model has been significantly improved compared with other baseline models.3.The prototype system of community question and answer is designed and implemented.The data collection module,data preprocessing module,question classification module and answer quality evaluation module are constructed.The prototype system can collect and preprocess the data,and classify the question sentences by the question classification module,and return the correct answer through the answer quality evaluation module.