Research On Course Answering System Based On Artificial Intelligence

The automatic question and answer system combines the artificial intelligence technology and natural language processing technology.On the basis of correct understanding of user semantics,the answer is simplified and returned to the user.Correct understanding of natural language will become one of the hot research directions.The automatic question and answer system is divided into limited and non-limited domains according to the research direction.As the limited domain direction of the automatic question and answer system,there are still many blanks in this field.Therefore,this paper studies the answer after the software test class,designs and implements the answer system.The main research contents are as follows:(1)Build a knowledge base for after-class software testing.This paper collects data in two ways,one is to use free and public crawler software to crawl data in Baidu Encyclopedia and forums,the other is to collect data in software test books and university test papers by manual methods,and then clean all collected data using Pandas.(2)Natural language processing.On word segmentation,use the Jieba thesaurus toge-ther with custom software to test the professional field dictionary to improve the accuracy of Chinese word segmentation;on de-stop words,use the Harvard University to remove words that are not meaningful to sentences;on key word extraction,select the improved TextRank algorithm to achieve keyword extraction;on synonym processing,use the Harvard synonym forest extension.(3)Semantic similarity calculation.In this software test and answer system,the core idea is to use natural language to query questions.This answering system first performs natural language processing,such as word segmentation,de-stop words,keyword extraction,synonym processing,etc.Then it uses CBOW model in Word2Vec to train word vectors,constructs word vector matrix,and finally uses word vector matrix as the input layer of convolution neural network.In the pooling layer,max-pooling is used to improve the characteristics,and the final output is similarity value,if the threshold value is exceeded,the query results are output in reverse order.(4)Implementation of Software Test Answer System.The system uses Flask to visualize.It has two roles:teacher and student.Teacher-side functions include querying questions,viewing all knowledge bases,hot questions,questions to be solved,my answers and personal information center.Student-side functions include querying questions,viewing all knowledge bases,adding questions not found,collecting questions for personal queries and personal information core...