Research On The Subjective Question Scoring Model Based On Case-based Reasoning

Because of the particularity of the domestic language,it is necessary to carry out the Natural Language Processing to deal with the subjective questions further.The automatic score of the subjective question belongs to the Natural Language Processing module.At present,the examination system in China can only score some objective questions,and can not score the subjective questions automatically.The subjective automatic scoring algorithms in China include the subjective score based on the vector space model,the subjective algorithm based on the unidirectional close degree,the sentence similarity calculation method based on the dynamic programming,and the automatic score of the Chinese subjective questions based on the improved LDA model.These algorithms do not take into account the logical relationship between the semantics.In practical application,the effect is not good.This thesis proposes a case-based reasoning based automatic scoring system for subjective questions.Case reasoning theory is applied to automatic scoring of subjective questions.The Chinese sentence is carried out in Natural Language Processing,the sentence is divided into different words by the word segmentation technology,and the logical relationship between words and words is considered.The dependency syntax is applied to case reasoning in the case retrieval stage,and the sentences are divided into key words,core words and other words.The main part of the sentence is found out,and the text similarity calculation model is separated from the word frequency.The logical relationship is taken into consideration,and the accuracy of the subjective score is further improved.In the case reuse stage,a case reuse model with confidence constraints is established,which is different from the traditional method.In this thesis,a method of dynamic determination of threshold is proposed.The reliability and accuracy of the result of case reasoning can be improved by optimizing the confidence threshold to improve the reliability and accuracy of the case reasoning.At the same time,the problem of the threshold value is too high.The Boot Strap method is used to amplify samples,and the threshold can be effectively determined for data with smaller sample size.At the end of the article,the architecture,business process and functional modules of the automatic scoring system are designed from the perspective of software development,and the design and implementation of a CBR Based subjective scoring system are completed.