Distilling Knowledge Learned in BERT for Text Generation
This work addresses the challenge of leveraging BERT's bidirectional capabilities for text generation tasks like machine translation and summarization, offering a novel method to enhance coherence in auto-regressive models.
The paper tackled the problem of adapting BERT for text generation by proposing Conditional Masked Language Modeling (C-MLM) to fine-tune BERT and use it as a teacher to improve Seq2Seq models, resulting in significant performance gains over Transformer baselines and achieving new state-of-the-art on IWSLT German-English and English-Vietnamese machine translation datasets.
Large-scale pre-trained language model such as BERT has achieved great success in language understanding tasks. However, it remains an open question how to utilize BERT for language generation. In this paper, we present a novel approach, Conditional Masked Language Modeling (C-MLM), to enable the finetuning of BERT on target generation tasks. The finetuned BERT (teacher) is exploited as extra supervision to improve conventional Seq2Seq models (student) for better text generation performance. By leveraging BERT's idiosyncratic bidirectional nature, distilling knowledge learned in BERT can encourage auto-regressive Seq2Seq models to plan ahead, imposing global sequence-level supervision for coherent text generation. Experiments show that the proposed approach significantly outperforms strong Transformer baselines on multiple language generation tasks such as machine translation and text summarization. Our proposed model also achieves new state of the art on IWSLT German-English and English-Vietnamese MT datasets. Code is available at https://github.com/ChenRocks/Distill-BERT-Textgen.