Research On Question-Answering Technology Of College Entrance Examination Volunteers

College entrance examination,as the last gateway to higher education,is an important turning point in life planning.Therefore,the selection of college entrance examination volunteers is crucial.Before filling in the college entrance examination,candidates should have a certain understanding of the information of various schools and majors,so as to avoid blind choice,which will bring troubles to the future university study and life.Today,with the explosive growth of information,how to quickly and accurately obtain useful school,professional and other information in a mass of information has gradually evolved into a more important subject.In recent years,the rapid development of artificial intelligence,has made a breakthrough achievement,more and more aroused people's attention and attention.An important branch of artificial intelligence is the question answering system.Questionanswering system is a more human and efficient information retrieval technology,which integrates natural language processing technology.This paper is a research on the question and answer technology for college entrance examination volunteers.Since few people do the research on the question and answer technology in this field,the research in this paper has certain practical value and significance for the future question and answer system construction in this field.This paper is based on information retrieval technology research for the college entrance examination volunteer question and answer system.Firstly,this paper collects relevant questions and answers of college entrance examination volunteers from various college entrance examination network platforms,college entrance examination APP platforms and Baidu zhidao through crawler technology,and stores the data with My SQL database.Secondly,the data set is preprocessed and manually classified according to the question and answer obtained by the crawler,so as to prepare the data required by the research experiment of question and answer technology in the future.In this paper,deep learning and the deep learning method of introducing attention mechanism are mainly applied.The specific research contents are as follows:1.Process the data acquired by crawler,manually mark the problem category and problem component,establish the problem classification corpus and extract the problem component corpus.2.Traditional machine learning method SVM is adopted for problem classification;Deep learning based methods CNN,LSTM;Two deep learning fusion methods BiLSTM +CNN;In the Bi-LSTM +CNN method based on Attention,a total of four models were used for text classification,and the classification effect of the model was evaluated through the macro average and micro average of evaluation indexes.The results of model experiment show that the classification method of Bi-LSTM +CNN questions based on Attention has higher accuracy than that of SVM,LSTM,Bi-LSTM +CNN,which is helpful to improve the accuracy of question classification.Its macro average value is 95.64% and micro average value is 94.47%.3.The problem components were extracted into sequence labeling problems.In this paper,CRF model and Bi-LSTM +CRF model were used for sequence labeling experiments.The model effect is evaluated by evaluating index F1 value.According to the results of model experiments,the effect of sequence labeling of CRF model is better than that of Bi-LSTM +CRF,which may be because the corpus in this paper is too small and the deep learning model cannot be well trained.In view of this problem,this article carries on the forecast,in the follow-up research work solves this problem.4.Build confidence sorting algorithm based on the deep learning answer,combined with the introduction of attention mechanism in a sentence and Bi-LSTM model to calculate the answer confidence,compared three models(CNN,LSTM,introducing the attention mechanism of Bi-LSTM)in a sentence on the answer confidence sorting results,you can know: the introduction of Bi-LSTM attention mechanism model,in a sentence to a certain extent,improve the accuracy of the answer choice,the model experiment,the best effect.