No abstract

Automated Essay Scoring is a very significant research subject for the processing of machine scoring. Computer-based College English Test (National English level test) further motivates the research of AES system for Chinese English leaner. In this paper, we introduce an effective AES system based on computer-based CET4. Features belong to three domains: language quality, content and organization, are involved in our system to figure out the feature collection with high correlation coefficient. The method of improved TF*IWF*IWF in computing term weight is used to optimize content-features. Through comparison of five scoring models based on classical text classification algorithms, we proposed an algorithm based on Adaboost, Voting and KNN to improve the accuracy of machine scoring. The exact agreement attends to 60%, and the adjacent agreement is above 95%. Moreover, we also discovered that writing patterns for Chinese English learners emphasize on fluency, lexical complexity features and two features from organization.

No abstract