Multi-Head Multi-Layer Attention to Deep Language Representations for Grammatical Error Detection

Abstract

It is known that a deep neural networkmodel pre-trained with large-scale data greatly improves the accuracy of various tasks, especially when there are resource constraints. However, the information needed to solve a given task can vary, and simplyusing the output of the final layer is not necessarily sufficient. Moreover, to our knowledge, exploiting large language representation models to detect grammatical errors has not yet been studied. In this work, we investigate the effect of utilizing information not only from the final layer but also from intermediate layers ofa pre-trained language representation model to detect grammatical errors. We propose a multi-head multi-layer attention model that determines the appropriate layers in Bidirectional Encoder Representation from Transformers (BERT). The proposed method achieved the best scoreson three datasets for grammatical error detection tasks, outperforming the current state of the art method by 6.0 points on FCE, 8.2 points on CoNLL14, and 12.2 pointson JFLEG in terms of F0.5. We also demonstrate that by using multi-head multi-layer attention, our model can exploit a broader range of information for each token in a sentence than a model that uses only the final layer’s information.