Research And Implementation Of Automatic Marking Subjective Questions

As we all know, the examinations in the teaching process are important meansfor a teacher to detect students’ knowledge level. Among them, subjective questioncan systematic investigate the mastery of knowledge of students, thus becoming animportant exam question. However, the marking work of subjective question arevery dependent on artificial. With the number of students who take the examincreasing, time-consuming and labor costs on marking subjective question are veryhigh. So the study of subjective questions intelligent marking technology has got agreat research value to be researched. Studying subjective questions intelligentmarking technology, this thesis are mainly related to the Chinese wordsegmentation, text similarity computing, and related technologies of the Chineseinformation processing technology.Based on the existing theory, major improvements of this thesis have beenmade in two areas:Firstly, this thesis proposed the post-processing algorithms of chinese wordssegmentation results, and improved the CRFs word identification rate on theOOV(Out-Of-Vocabulary) word. The post-processing algorithms in this thesis arebased on the semantic role labeling results of sentence.Secondly, this thesis implemented the automatic marking system of subjectivequestion, and the marking service was packaged into callable interface. This thesismainly determines the similarity of the answers text and the standard answer textthrough three aspects, including word form, word order, and semantics. This thesisalso improves the calculation method of word order similarity with adding the wordsemantic similarity calculation.This thesis considers the evaluation of automatic marking technology ofsubjective questions as this way: The ideal state of automatic marking technologyof subjective questions is that, the marking result difference between the automaticmarking technology and artificial intelligence marking scoring is within10%orless. The experimental results of this thesis show that with the post-processingalgorithm of this thesis, the identification level of OOV word has been imroved,and word’s segmentation effect has also been improved. Meanwhile, this thesis notonly subjective questions presented feasible solution of the automatic markingwork of subjective question, but also conducted test experiments. The experimentalresults show that results of this thesis, have some practical value in the field ofautomatic marking technology of subjective questions.