Research On Extracting Key Evaluation Features Of Chinese Compositions In Primary School

The second and third stages of primary school should have the basic ability of"writing",which is the requirement of Chinese curriculum standard for writing.The standard also puts forward that students' writing level should be comprehensively investigated according to the requirements of different stages.At present,the primary school composition evaluation still uses the method of teachers' subjective scoring,which is time-consuming and labor-consuming,and has strong subjectivity.It also affects the reliability and validity of composition evaluation.Based on this,the research on automatic composition evaluation technology has been carried out at home and abroad.However,at present,the research of automatic composition evaluation technology has not focused on the writing level of the second and third primary school students whose mother tongue is Chinese.Moreover,the existing various scoring systems only select limited evaluation features,not being comprehensive analysis to the composition text.In view of the above situation,this thesis analyzes the evaluation features of the second and third primary school compositions from the three levels of lexicon,syntax,and chapters,obtains the comprehensive evaluation features,and realizes the automatic extraction of key evaluation features.The main work includes:Firstly,this thesis analyzes the research status of the composition automatic evaluation technology at home and abroad.And it is found that the current research in China is not combined with the requirements of the primary school curriculum and the teaching practice,and there is lack of guiding significance for composition teaching evaluation in the primary school.Based on this,this thesis uses Coh-Metrix Chinese text analyzer to comprehensively analyze 106 lexical grammatical and semantic features that have significant effect on text distinction.This thesis obtains 4 comprehensive evaluation features that have the highest regression model parameters with the composition scores,which including 2 evaluation features under the lexical level,1 under the syntactic level and 1 under the textual level.Secondly,it discusses the automatic extraction of the syntactic coherence evaluation feature at the syntactic level.The description system of syntactic coherence features is obtained through literature research,which is the theoretical support of automatic extraction method.By designing the self-defined dictionary,this thesis perfects the presentation of 11 kinds of complex sentence structure.Besides,it employs the jieba technology to realize Pos-Tagging and python programming language to realize code-designing.These work realize the automatic extraction of evaluation features under the syntactic level,and the experimental results are presented in a visual interface.Through reliability test and correlation analysis,the validity of the automatic extraction method of evaluation features at the syntactic level is verified.Finally,this thesis discusses the automatic extraction of the evaluation features at the textual level.This thesis analyzes the basic algorithms of Chinese text keyword automatic extraction,and adopts TextRank algorithm to realize the automatic extraction of text keyword.By building thesaurus,calculating the weight value of the correlation between keywords and text topic,this thesis realizes the automatic recognition of the text topic-correlation evaluation characteristics.Through the correlation analysis,the effectiveness of the automatic recognition method of topic correlation evaluation features is verified.Based on the daily composition and examination composition in the second and third learning stages of primary school,combined with the actual situation from the teaching in classroom,this thesis makes a comprehensive analysis of the automatic composition evaluation features and obtains the comprehensive evaluation characteristics.This thesis discusses the automatic extraction of key evaluation features,obtains the automatic extraction method of key evaluation features,and verifies the feasibility of the method.This thesis provides a reference for the development of automatic composition evaluation in primary school.