| In recent years,artificial intelligence technology has been widely used in the education industry,such as intelligent paper composition and automatic scoring of subjective questions.Due to the fact that it takes a lot of time and energy to correct subjective questions,and the score is easily affected by the teacher’s subjective consciousness,Moreover,natural language processing technology in artificial intelligence is becoming more and more mature,so it is a future trend to use natural language processing technology to realize automatic scoring of subjective questions,which is of great significance to promote the intelligent development of the education industry.The realization of automatic scoring of subjective questions,the key technology is to calculate the semantic similarity between texts.For the realization of calculating the semantic similarity between texts,a large number of algorithms have been proposed,among which the LSA algorithm is one of them.On the basis of the LSA algorithm,in view of the current problems of LSA processing Chinese,several improved LSA algorithms are proposed for the calculation of the semantic similarity of Chinese text and the automatic scoring of subjective questions,and finally the algorithm that can handle Chinese subjective questions is applied Based on the automatic scoring system for subjective questions,the automatic scoring of subjective questions is realized.The main work is as follows:1.In view of the problem caused by the LSA algorithm processing Chinese text and ignoring the text grammar,resulting in a large difference between the obtained semantic similarity results and the real semantics,an improved LSA algorithm(e LSA)is proposed,that is,the LSA algorithm combines dependency syntax and corresponding rule calculations Semantic similarity between texts.Experimental results show that the semantic similarity results calculated by the improved LSA algorithm are closer to reality than the original algorithm.2.Since the above-mentioned improved LSA algorithm can only deal with relatively simple Chinese texts at present,a new way of thinking is needed for processing texts with more complex subjective questions,so an improved LSA algorithm(DP-LSA),the original LSA algorithm,the greedy algorithm I implemented myself to calculate the semantic similarity algorithm between texts and the algorithm to calculate the semantic similarity using subject-verb-object keywords The difference between the number of sentences is different and the corresponding algorithm is selected to calculate the similarity between the text semantics,and finally the similarity score is multiplied by the corresponding weight to obtain the score of the subjective question.The algorithm has been compared with LSA algorithm,TF-IDF algorithm,BERT and other models.The experimental results show that the DP-LSA algorithm shows the best performance in various evaluation indicators.3.In addition to proposing two improved LSA algorithms,a subjective question automatic scoring system based on DP-LSA algorithm is also developed.The users of the system mainly include teachers and students,the functions of teachers mainly include login,upload test questions,publish test questions,automatic marking papers,manual marking and grade analysis functions,and student functions mainly include login,registration,online question making and grade query and other functions.Teachers can realize the combination of automatic marking of subjective questions and manual marking of papers,or use one of them,which greatly improves the efficiency and accuracy of subjective questions,and the system test shows that the system has shown good results in realizing automatic marking of subjective questions in designated subjects,which provides certain reference value for the research of automatic marking of subjective questions. |
