Research On The Application Of Text Similarity Algorithm In Automatic Scoring System

The traditional way of examination-mark generally use manual scoring,which is a very time-consuming and inefficient scoring method.With the development of computer technology,computer automatic marking has been applied in the online examination system.Questions of the general course examination are mainly divided into subjective and objective questions,as the objective problem has a fixed answer,so the objective scoring scoring is not a technical problem.On the contrary,the subjective questions generally only give a reference answer,if you want to score for the subjective questions,the system needs to use Natural Language Processing as a technical route.This examination system takes the university public class examination as the research object to design the subjective question automatic marking system.According to the subjective questions scoring process,this system is mainly divided into the following parts: text segmentation,the prepose work to function words,synonyms,stop words,passive conversion)and orientation analysis,similarity matching calculation.Among them,we segment the sentence to by using the Thulac segmentation toolkit which is developed by the Tsinghua University segmentation of the text or sentence similarity matching is used to calculate the matching degree calculation VSM algorithm and TF-IDF algorithm.This system focuses on the similarity matching and tendency analysis and calculation,the thesaurus establish of the general courses like Marxist Philosophy and Ideological and Moral Cultivation based on the examination system to make it more suitable for text segmentation.After this,the words substitution is used in order to increase the reliability of the similarity calculation.We compare the standard answers and the student answers circularly before the similarity matching.If the opinions of the two sentences with the tendency to express the views are completely opposite,it may not be necessary to calculate the similarity at all.VSM and TF-IDF are wrote on the method of similarity calculation,finally we use the similarity calculating algorithm which is based on Word2 Vec.