Research On Automatic Scoring System Of Subjective Questions Based On TF-IDF And LSI

In China,the examination has become an essential part of Chinese daily life.It is as small as learning a unit or a chapter of tests,weekly exams,monthly exams,mid-term exams,final exams and so on.However,in addition to computer-related examinations,other types of examinations can currently only perform automatic scoring of multiple-choice machines.Fill-in questions,short answer questions,essay questions,etc.still need to be manually scored,and the reviewers are usually required to be in very short time.Reviewing the thesis will undoubtedly bring great workload to the reviewers,and has severely reduced the efficiency and accuracy of marking.Therefore,the problem of how to automatically score short texts such as fill-in questions and short answer questions is achieved by using existing computer technology to measure text similarity with standard answers.First of all,the research status of the automatic scoring system for subjective questions at home and abroad was reviewed.The subjective questions of the system modeling and simulation test for industrial engineering of Kunming University of Science and Technology were used to apply the automatic scoring system.Secondly,the programming language Python was used for programming.The jieba Chinese word segmentation tool was used for text preprocessing.The TF-IDF and LSI text similarity measurement methods were used as the text analysis model.The subjective questions based on TF-IDF and LSI were automatically scored.Then,on the basis of the system modeling and simulation course,30 junior students were randomly selected to answer questions,and the results of the answer were systematically scored and scored by teachers,and the score results were compared and analyzed.The analysis results showed that the consistency between TF-IDF model scores and teacher score results are all above 83%,while the consistency between LSI model and teacher score results is below 30%.Therefore,the system under TF-IDF model has better applicability.In this thesis,based on the experience and characteristics of teacher's rating,an automatic scoring model for subjective questions based on TF-IDF and LSI was established,and an automatic scoring system based on the model was implemented,and examples were verified.The verification results showed that the scoring effect of the TF-IDF model in this system was close to the actual teacher score,and had certain validity and practical value.At the same time,in this thesis,the TF-IDF and LSI-based subjective topic automatic scoring system developed can automatically score students' answers.This laid the foundation for future scholars to conduct research on subjective topic automatic scoring.