Research On Classification Model Of English Grammatical Error Correction

With the increasing popularity of English all over the world,more and more attention has been paid to the study of automatic English grammatical error correction(GEC)in academia and industry respectively.The algorithm with strong error-correcting ability can help people reduce grammatical errors and improve the efficiency of life and production in the process of learning and using English.We treat grammatical error correction as a classification problem in this study,where for different types of errors,a target word is identified,and the classifier predicts the correct word form from a set of possible choices.We propose a novel algorithm for GEC.Firstly,when defining the input of the model,we locate the target word in the sentence according to linguistic information to extract the context of the target word.Secondly,we use RNNs with two attention mechanisms to compute the feature representation of the target word context,and use the multi-layer perceptron to select the classification label according to the information obtained.By calculating and processing a large number of free plain text on the Internet,we can get training data with labels,which solves the problem that the current supervised English GEC algorithm lacks a large number of labeled data.Finally,our model is trained in an end-to-end fashion.Experiments are carried out on the CoNLL-2014 test set to calculate and analyze the precision,recall and F0.5 of the algorithm.Experimental results show that our novel approach outperforms other classifier methods on the CoNLL-2014 test set by a large margin(F0.545.05%vs.41.6%),and achieves the state-of-the-art.