An Imputation Bi-directional Long Short-Term Memory for Detecting English Grammar Error based on Machine Translation

The Machine Translation has enormous application and better growth prospect. The English learning is significant part in which grammar has difficult problem for learners to tackle the learning procedure. In this research, the Imputation Bi-directional Long Short-Term Memory (Im-BiLSTM) is proposed for detecting English grammatical error correction in machine translation. The N-gram model is used in this work for preprocessing the English grammatical errors which contain two types; both first and second types are used to judge if the grammar is accurate or wrong. The Im-BiLSTM is straightly learn from sequence information which efficiently reorganized if error transfers at backpropagation. The gate mechanism defines the BiLSTM adaptively which recollects past state and learns present data. The f1-score, precision, accuracy and recall are used for evaluating Im-BiLSTM performance. The Im-BiLSTM achieves 98.91% f1-score, 98.89% precision, 99.82% accuracy, 98.37% recall when compared to ASS-TG and Hybrid English error Detection and Correction (HERDC).