A correcting method for article error in English essays of Chinese students based on hybrid features classification

Abstract

This paper presents a method for automatic article identification and correction in writing by Chinese learners of English. We train on one million noun phrases extracted from a corpus of published textbooks. In extracting features, we use the n-grams for local context features in the form of words and part of speech tags and the parse tree for syntactic features which is more linguistically sophisticated. At the same time, this paper raises a new approach based on mutual information and the contribution for training and classification. Performance of this new approach shows both effectiveness and efficiency, and the results are a significant improvement on the previous best results. © 2013 Springer-Verlag.