The Key Technology Research On Automated Essay Scoring

With more and more high school students to take part in the entrance examination of American University(SAT),writing as an important part of the SAT test,has an important influence on the exam scores.However,for domestic non-English speaking senior high school students,English writing is always a weak point.They have urgent need to improve it.In the traditional English writing teaching,because the weakness of teachers correcting composition for a long time,the feedback is not timely and the effect is not obvious and the correcting process is vulnerable to other human factors,the effect of students' writing training is far to meet the requirements of the examination.Automated English essay scoring system is a tool of automated scoring using computer science and natural language processing technology.It presents students' writing error visually after scoring to help the optimization of students' writing skills and overcome the above drawbacks thus makes teachers and students benefit and improve the level of English writing.Currently foreign countries have mature scoring system,and the actual use is the pretty good,but these systems are designed and developed for English learners,and are not very suitable for non-English speaking students.So for the English learners,the scoring strategy of system should change and the feedback should improve.This paper is to test SAT online lerning platform as the basis,in view of the SAT writing features of the candidates to effectively explore the characteristics of items reflect the writing quality,and on this basis,establish a reasonable assessment model.This paper feature analyzes and extracts the students' compositions on the platform based on natural language processing techniques and statistical regression analysis,and predicts scores of composition by means of characteristic dimensions and rating analysis model,detects error in the marking process and displays the marking result visually.This subject extracts vocabulary,collocation,grammar,paragraphs and other characteristics from 572 students' compositions,mainly analyzes on five score module(the use of vocabulary,grammar,semantics,theme arguments and textual structure)to extract characters related to it.Then apply to other scoring module feature analysis by analogy of this.And researches visualized display grading results of the composition to display different types of classification errors which will appear in the composition so that to remind students the errors.Finally the positive influence and possible problems of review information visualization on teaching are briefly analyzed,and summarizes the advantages and disadvantages of the article,and looks forward to the research goal and direction in the future.