Automatic Scoring Algorithm Of Subjective Questions Based On LDA Topic Model

With the rapid development of computer science and technology,traditional subjective examination has been transferred from written examination to computer,and thus automatic scoring technology has attracted great attention.At present,the computer as an objective scoring machine is already a mature technology,which is used for automatic scoring in objective topic(fill in the blanks,true or False,choice),and improve scoring efficiency and accuracy.However,the computer automatic scoring technology is not very mature,especially in the field of Chinese subjective questions.Automatic scoring of subjective questions is not only a simple process automation but a theory and technology with natural language processing and pattern recognition etc.Since the diversity of Chinese sentence structure and the form of expression,Chinese subjective automatic scoring develops slowly.The application of automatic scoring technology in the subjective field,can be a good solution to the manual marking by the external factors,such as the degree of neatness of the candidates,the mental states of the staff,the degree of normative font of the candidates,thus effectively guarantee the fairness and impartiality of competition between candidates.The computer is fast and efficient,and it is not influenced by the subjective factors,which are very suitable for the automatic scoring of subjective questions.Therefore,it is of great significance to apply the computer to the subjective subject automatic scoring field.This paper examines the existing domestic automatic scoring system,summed up the existing domestic subjective problem automatic scoring method.At present,the more commonly used method is to compare the similarity between standard answer and the candidates.In the similarity comparison,it based on dynamic programming,based on keywords,based on syntax and so on.As the complexity of Chinese language and the characteristics of subjective questions,these methods also have some room for improvement in spite of its certain effects.This paper proposes an automatic grading algorithm based on the subjective problem of LDA topic model on the basis of summarizing the existing automatic scoring technology and considering the characteristics of the standard answer and the way of marking the scoring staff.The algorithm preprocess the text of the subjective question first,and then improve the text.We use the LDA(Latent Dirichlet Allocation)topic model to deal with the subjective questions and get the corresponding theme distribution,by comparing the standard answers and the similarity of the subjects',so as to achieve the purpose of subjective questions automatic scoring.The experimental results show that the algorithm has good practicability and accuracy.At the end of this paper,summed up the algorithm,point out the shortcomings,and looking ahead for future work.