Explore Deep Learning for Chinese Essay Automated Scoring
Haojin Li, Tao Dai · Journal of Physics Conference Series · 2020
Abstract Automated Essay Scoring (AES) has gained increasing attention in recent years. In this paper, we propose a novel automated Chinese essay scoring model, called BLA (BERT and Bi-LSTM with Attention), by using neural network. The model uses a BERT network to obtain the sentence vectors for an essay, and then uses a Bi-LSTM network with two layers to extract the essay vector. We also consider topic instruction as a long sentence and obtain its vector representation by BERT. The obtained topic instruction vector is then used as attention information to further help obtaining more effective essay vector. Moreover, we also present a new dataset that contains topic instruction and related essays collected from Chinese high school. The detailed experimental results show that our proposed model outperforms other baseline methods.