The Design Of Automatic Feedback Model For Chinese Academic Comments

With the development of artificial intelligence in education,many scholars began to pay attention to the use of computers to assist teachers in evaluation of large number of students' academic reviews to help students improve their academic writing.In previous studies,some scholars have verified that peer review can help students improve writing;compared with offline peer review activities,online peer review activities are more potential in improving the effectiveness and reliability of peer review.However,as student reviewers are unable to provide appropriate comments,student writers cannot easily capture and apply the comments received.Therefore,based on the online peer review environment,this study aimed to improve the peer reviewer's review quality during the peer review process,by developing a system that automatically evaluates student Chinese reviews and provides meta-review feedback for reviewers,helping reviewers to make better reviews so that they can reflect on themselves and write better and more effective reviews.The main research work is as follows:Firstly,for online peer review activities,this paper proposes a review automatic feedback model,which includes three review quality indicators:review content type,review relevance,and comments feedback.The automatic comment feedback model takes automatic classification of comments as an entry point and quantifies the three comment quality indicators with the help of natural language processing,machine learning,and deep learning technologies.The content types of reviews are discussed from three dimensions:affective,cognition and metacognition.In the 4,652 peer review data collected and manually marked,the performance of automatic classification of reviews using LSTM classifiers is higher than that of traditional machine learning reviews:In the cognitive dimension,the accuracy of the LSTM classifier is 76.2%,which is higher than Naive Bayes classifier(69.8%);in the metacognitive dimension,the accuracy of LSTM classifier is 81.7%,which is higher than Naive Bayes classifier(76.8%).Secondly,the author developed the "online writing-peer review feedback" system based on the comment automatic feedback model.The system can provide interfaces such as online writing,peer review and system feedback,it also implements the function of anonymously reviewing articles by users,which can ensure the fairness of the review.The system can automatically generate comment feedback after peer reviewers submit comments,and display it on the system's comment feedback page.Thirdly,this research cooperates with the actual postgraduate academic writing courses to allow postgraduates to complete the online peer review process on the system.Each postgraduate is both a writer and a reviewer of the article.Through the analysis of the review data before and after feedback in the system,the effectiveness of the automatic feedback mechanism of reviews on improving the quality of student reviews is explored.The results show that:the automatic feedback mechanism of reviews has a certain improvement on the quality index of the relevance of comments of 30 graduate students(0.089).In addition,after the feedback,the total number of words and the number of comments of the 30 graduate students have increased,indicating that the feedback process can improve the willingness of the graduate students to write comments.