Research on English Grammar Error Correction Technology Based on BLSTM Sequence Annotation

Abstract

Grammatical errors are one of the most common types of errors in English learning and writing. Intelligent detection and correction of English grammatical errors can effectively help English learners improve their learning efficiency. The research is based on the BLSTM bi-directional and long-short-term memory neural network to construct a sequence labeling model to detect and correct English grammatical errors. The results show that the sequence tagging model based on the BLSTM bi-directional and long-short-term memory neural network has an accuracy of 96.88% for the part-of-speech tagging of special corpora containing grammatical errors, the accuracy rate of error correction is 33.58%, the recall rate is 44.95%, F-measure index value is 38.26%, which is 4.85% higher than the UIUC model and 4.92% higher than the Corpus GEC model. It can automatically detect and correct text grammatical errors with good effect, which provides a new idea for the development of intelligent English grammatical error detection and correction technology.