Research On The Automatic Classification Evaluation Method Of Compositions Of Primary School Based On Rhetoric Use

With the rapid development of information technology,students have been able to use the computer and network for independent learning of composition,and the automatic evaluation of compositions is an important part to support students'extracurricular independent learning of compositions.At present,there are many researches on the automatic evaluation of compositions,and there are also some mature automatic scoring platforms for compositions.However,almost all of them are aimed at the automatic marking of college entrance examination papers and English compositions.In order to solve this problem,this thesis studies the automatic evaluation method of primary school composition for the needs of informationization teaching in primary school Chinese.Considering that there are more types of compositions for primary school students than those for college entrance examination,generally,there are writing people,writing scenery,writing things and writing objects and the evaluation standard of primary school composition is simpler and clearer than that of college entrance examination composition,therefore,this thesis studies and proposes an automatic classification evaluation method of primary school compositions based on rhetoric use.The main research work of this thesis is as follows:(1)This thesis analyzes the relevant current available technologies,analyzes their basic principles and characteristics,and establishes the basic research program.(2)TextRank combined with character-level convolutional neural network is used to realize the automatic classification of primary school compositions.First,the TextRank based keyword extraction algorithm was used to remove some redundant semantic information from the sample text,then the application of Word Embedding text representation of data set as a convolution of the neural network input,through iterative training and testing,the model is finally implemented.Experiments show that this method can improve the performance of composition classification significantly.(3)This thesis studies the automatic extraction of Parallelism,figurative,anthropomorphic and quotable sentences in primary school compositions.By analyzing the characteristics of all kinds of rhetorical sentences,specific implementation plans are put forward.The automatic extraction method of parallel sentences is based on rule-making.The automatic extraction method of figurative and anthropomorphic sentences is based on deep learning.The automatic extraction method of reference sentences uses the method of establishing reference library.These methods show good results on dataset testing.(4)Using detailed data sets and training different machine learning classifiers for comparison,an automatic classification rating system for primary school compositions based on rhetorical use was finally obtained.At the same time,the corresponding comments were designed according to the characteristics of each scoring feature.Finally,the thesis integrates all the previous research modules to obtain the primary school composition automatic classification evaluation system based on rhetoric use and the correctness of the method is verified by comparison test.