Research And Implementation Of Simulated Platform And Intelligent Scoring System For College Entrance Examination Of English

The development of artificial intelligence makes people's life and work easier and more effective.Online examination and scoring system,which is based on computer,not only improves the efficiency of students' learning,but also reduces the pressure of teachers' evaluation work.The evaluation of objective questions has experienced from manual evaluation to cursor reader evaluation to computer character matching,and the correct rate of evaluation has soared to 100%.For subjective questions,foreign countries have experienced PEG,E-rater and other systems,while domestic systems using similarity comparison of large English corpus,systems based on natural language understanding and intelligent evaluation technology,etc.Most of these systems are based on some shallow linguistic features such as rules and LSA.They have no deep perception of English language sense.Although the current intelligent scoring system has made a lot of achievements,they have not fundamentally solved the rationality of subjective intelligent scoring.In order to perceive the sense of language better in English compositions and improve the rationality of intelligent assessment,this paper proposes an N-language sense quantification algorithm based on relevance analysis and an English composition score fitting algorithm based on rationality enhancement.Quantification of language sense value calculates its support in corpus by acquiring the N-tuple of the composition.If the degree of support is below the threshold,then the part with language sense problems will be analyzed and the types of language sense problems will be distinguished to help students to revise their compositions.In addition,this paper also extracts the features of words,sentences and texts to fit the scores of English compositions.Since not all students can complete their compositions according to the requirements of the topic,the shunt scoring model is adopted to separate the normal and low-grade compositions.Low-grade compositions are scored by k-nearest neighbor algorithm,while normal compositions are scored by support vector regression algorithm.It is found that the composition scores also show a certain normal distribution through statistics.Standard Support Vector Regression(SVR)algorithm is prone to data skewness,so this paper adopts the method of rationality improvement to solve this problem,and gives the corresponding penalty factor according to the distribution of data sets.Finally,a simulated intelligent scoring model of college entrance examination composition is produced by training the simulated test data of a high school in the past years.By testing the test data set,the experiment shows that the extraction of the language sense features of the candidates' English compositions can not only judge whether the candidates' language sense is problematic,but also provide a basis for the overall evaluation of the composition.The proposed algorithm can improve the prediction accuracy of some data and solve the problem of data skewness with normal distribution of scores.This paper designs and implements an online simulation examination system,which integrates the composition intelligent scoring system to meet the requirements of teachers and students on-line examination and intelligent evaluation.