Design And Implementation Of Evaluation System Of Chinese Essay Of High School Students Based On Rhetoric Sentences Recognition

As an important assessment task in all Chinese language exams,the evaluation of essays is mostly carried out manually,which is often labor-intensive,and the process of essay evaluation is often unavoidably subjective so as to be unfair.In that case,the above problems can be solved to a certain extent if the evaluation of essays can be done automatically by machines.More importantly,if we can give specific comments after scoring,it will be highly beneficial to the improvement of students writing skills.In this paper,Deep learning and natural language processing techniques are adopted to automatically identify rhetorical sentences in Chinses essays of middle school students,and hereby some simple comments can be generated.Regression models ware also trained to gives the corresponding overall score.Finally,this paper designs and implements an automatic evaluation system for Chinese essay of middle school student automatic evaluation system based on rhetorical sentence recognition,which has been tested to work well.For the automatic recognition of parallel sentences,this paper integrates various features,including the input sentences’ token sequences,POS sequences,and emotional polarity sequences.The interaction operation is made to obtain the interaction matrix respectively,then feeds in the CNN in the form of multiple channels to extract the local features layer by layer.Finally,the probability of whether two sentences belong to parallel sentences or is given,and through merging process the whole parallel sentences in the composition can be extracted.For the recognition of simile sentences,LSTM is used to encode two segments of the sentences divided by the indicators,and the attention mechanism is utilized to capture the connection between the tenor and vehicle inside the two segments,which can effectively extract the simile sentences in the students’ essay.The recognition of quoted sentences mainly relies on the construction of a quoted resource library.Elasticsearch is used to store these quoted original text because of its advantage of internal inverted index structure,and then a similarity method is conducted on the basis of the longest common substring between potential sentence and the quoted original.Finally,all the extracted rhetorical sentences features are viewed as a feature to integrated into a baseline learner only involving some shallow linguistic features.Eventually a regression model is trained by testing four regression algorithm,while the effectiveness of rhetorical feature is analyzed.All methods were verified on the dataset and showed good results.For the system implementation,this paper performance calls to multiple algorithm modules for feature extraction of students essay in a multi-threads using Python,realizes an automatic essay evaluation system for middle school students,and provides some simple remarks as feedbacks.Now it has been deployed and trial run in the server.