Design And Implementation Of Question Answering System For Programming Course Based On Frequently Asked Questions

With the development of Internet technology,more and more students tend to use information from the Internet to make up for the shortcomings of knowledge.But the search engine has the problem of information redundancy,which is not conducive to knowledge retrieval in specific fields.Therefore,it is important to efficiently screen out the most needed information.The course question answering system is the key technology to solve this problem.The system can help students solve the problems encountered in the course,so that they can quickly obtain the answers to the questions,which has important research value and application value.In this dissertation,a course question answering system was designed and implemented for the C programming course.Its main functions are as follows:analyzing the question text entered by the user to form a unified representation,classifying the multi-label text,and then searching in the corpus subset of the classified category,and returning to user the answer which question has the highest similarity comparing to user's question.In text classification,this dissertation proposes to train a classifier combined with the information of the answer summary in the corpus to make full use of the information in the corpus.In the similarity calculation,this dissertation proposes a word order sensitive similarity algorithm based on edit distance,which can identify the inconsistency between the query sentence and the key sentence.Firstly,C language question and answer pairs were collected,and the domain dictionary and the synonym dictionary for C language were constructed to form a basic corpus.In text classification,to make full use of the corpus,five training corpora such as question,answer,answer summary,question+answer,question+answer summary were used to train the classifier,and the performance of the model on different corpora was observed.At the same time,support vector machine,logistic regression and naive bayes classifiers were used as weak classifiers of the integrated model to improve the performance of the classification model.Secondly,in the process of comparing user questions to the questions in the corpus,for the problem that cosine similarity cannot pay attention to different word order,the original cosine similarity was improved based on the edit distance to suppress the similarity value of different word order sentences.Finally,a C language course question answering system was designed and implemented based on the framework Flask.The system is divided into front-end and back-end.The front-end includes question answering module and background management module.The back-end includes question understanding module,question classification module and answer extraction module.After testing,the function of each module of the system has reached the expected effect.Experiment and test results show that in the multi-label classification of questions,the scheme that combining the question with the answer summary proposed in this dissertation is more effective than only using question,answer,answer summary,or question+answer.The classification accuracy on the test set is increased by 1.33%,2.76%,2.47%,and 1.72%,respectively.In matching and positioning for user's question,the word order sensitive algorithm proposed in this dissertation has good experimental results on the test set,which can effectively suppress the similarity value of sentence pairs with different word order,and improve the accuracy and recall of question matching.All the functions of the course question answering system developed in this dissertation have passed the test,which can well solve the difficult problems of students in the learning process.