| Automated Essay Scoring(AES)is the evaluate and score of essays using techniques such as natural language processing,statistics and artificial intelligence.Its score is objective and fair,and don't need-a lot of manpower and material resources.At present,the widely used JuKu,BingGuo Intelligent Essay Evaluation System in China is not ideal for evaluating the off-topic essay,and the relevant foreign technologies and effects are unpublished,we can't verify its usefulness.Therefore,it is of great significance to research and implement a system that can accurately score the off-topic essays.This paper studies and implements an Automated Essay Scoring System for English essays,and focuses on the evaluation of off-topic essays,since there is no public data set specifically for the off-topic essay,Based on the Kaggle AES data set.We built a data set for the off-topic essay,then supplemented a part of the content enrichment.but irrelevant to the subject,and score it 0-2 points according to its writing level,This data set contains 1500 labeled samples.We proposed a simple and easy method to extract the topic vocabulary of essay,On this basis,it is used as the cosine similarity between the topic vocabulary vector of the essay and the vocabulary vector of the essay to be tested to measure whether the essay is off-topic.Then it analyses the shallow grammatical features of the essay,such as words,sentences,punctuation,etc.,further analyzing the features related to the semantics of the essay,such as part of speech,N-Gram,training the model with the tagged corpus,using this model design AES system for English essays.This system is simple and easy to use,and can accurately and quickly score English essays.We conducted detailed experimental tests on this trained model and this designed scoring system.The test results show that the model can effectively identify off-topic essays,and the scoring results are highly correlated and consistent with the manual scoring.This system is used to grade single(topic-specific)essay,batch essays and off-topic essays respectively.The results show that the system has good performance. |
PDF Full Text Request