Research On Automated Essay Scoring Method For Junior High School English

Writing occupies a large proportion in English exams.At present,the main way of scoring English essays is manual.Although it is scored by professional English teachers,there are still some shortcomings in the manual scoring,such as consuming a lot of manpower and material resources,strong subjectivity and big error.It is easy to be influenced by teachers' personal preferences and evaluation criteria,and feedback time for a long time.Compared with manual scoring,automated essay scoring has the advantages of objectivity,fairness,low cost,timely feedback and so on.Therefore,it is of great significance to study automated scoring of English essay.Traditional automated essay scoring methods mainly extract the shallow lexical,sentence and semantic features of the essay manually,and use machine learning model to score the essay.The workload of feature engineering is very heavy,and the finer-grained features of the essay are ignored.In recent years,with the development of natural language processing technology and deep learning,it has been applied to various tasks in the field of natural language processing and achieved remarkable results.Therefore,in view of the shortcomings of the traditional automated essay scoring method,this paper will construct an English essay scoring model according to the scoring criteria of junior high school English essays.Based on this model,an automated scoring method of English essays based on Hybrid Neural Network on Transformer Encoder(HNNTE)is proposed.This method uses the transformer encoder model to mine the shallow language information,syntactic information,semantic information and topic relevance information of the essay,so as to construct a more perfect automated essay scoring method to provide users with objective and fair scoring results.The main contents of this paper are as follows:(1)This paper studies the construction of a scoring model for English essays.Starting from the scoring criteria of English essays,this paper studies the features of essays that affect the scoring results,and constructs an automated scoring model of essays which includes shallow linguistic features,syntactic features,semantic features,topic-related features,so as to evaluate the quality of essays more comprehensively.(2)An automated scoring method based on HNNTE is studied.The traditional method of automated essay scoring is to predict the essay score by extracting the shallow linguistic and semantic features of the whole essay.In the actual manual scoring,the factors that need to be considered are far more than these.The deep semantic information,logical coherence of the content,and whether the essay is off-topic are also important aspects to be considered when scoring.In this stage,according to the scoring model,the shallow language features,syntactic features,semantic features and topic-related features are extracted respectively.The shallow language features are extracted mainly based on artificial rules,and other features are automatically constructed based on neural networks.In the process of model training,we use HNNTE model to test ASAP,an open data set of essay score,and compare the HNNTE model with the existing baseline method in the quadratic weighted kappa coefficient evaluation index.The experimental results show that the automated scoring method based on transformer encoder is better than the scoring method based on recurrent neural network.(3)Design and implement an automated essay scoring system.Based on the automated scoring method of English essays proposed in this paper,an automated scoring system for English essays is designed and implemented,which can provide real-time multi-dimensional score feedback for users and help them improve their writing ability.