Research And Implementation Of The Key Technology Of English Automatic Essay Scoring

English Writing Examination can test the comprehensive language use ability of English learners.In normal learning,teachers usually make manual evaluation,which is highly subjective,intensive and concentrated,resulting in low efficiency and high misjudgment rate.The development of computer technology has sped up the process of research and application on AES(Automatic Essay Scoring)technology,this technology can analyze the essay content and score,compared with the manual review,computers cost low,but higher efficiency review,such as spelling and grammatical errors.AES technology can also recommend advanced vocabulary and excellent essays,thus providing learners with more scientific essay writing guidance.However,most of the feedback of the current AES system is statistical information based on the essay content,and the evaluation of the quality of the logic and sentences is still not in-depth and accurate enough.Therefore,we'll not only improve the accuracy of the composition scoring but also evaluate the essay comprehensively.Thus making AES system better applied to the actual essay scoring.This paper studies the technology of AES,designs and implements the model of extracting features and discuss the ensemble approach.Firstly,this paper build the AES model,making a comparative simulation experiment and analysis of the key technologies.Then,introducing the graceful sentences recognition model.Through the research,we can determine the standard for evaluation of the gracefulness and then gather a lot of raw data on the Internet.We design the scheme of extracting features and integrating into the convolution neural network to train graceful sentences recognition model,we can use this model to quantify the gracefulness of each sentences and extract the sentence's graceful features by statistically quantified values,integrating them into the automatic grading composition.This paper introduces the concept of thematic relevancy feature,and calculates the similarity between the essay and its topic from three levels:word,sentence,and text,so as to represent the degree of relevance of the essay.Discuss the stacking ensemble method in order to improve the accuracy of AES model.Finally,a prototype system for AES is developed and implemented.After testing,each module of the AES prototype system can basically realize its function,the robustness and real-time performance of the algorithm have reached the basic goal,and can provide comprehensive evaluation information of the composition for learners and teachers,so as to improve marking efficiency.