Fine-tune BERT for Extractive Summarization
This work provides a performance boost for NLP researchers and practitioners working on summarization tasks, but it is incremental as it adapts an existing model.
The authors tackled extractive summarization by fine-tuning BERT, achieving state-of-the-art results on the CNN/Dailymail dataset with a 1.65 ROUGE-L improvement over previous methods.
BERT, a pre-trained Transformer model, has achieved ground-breaking performance on multiple NLP tasks. In this paper, we describe BERTSUM, a simple variant of BERT, for extractive summarization. Our system is the state of the art on the CNN/Dailymail dataset, outperforming the previous best-performed system by 1.65 on ROUGE-L. The codes to reproduce our results are available at https://github.com/nlpyang/BertSum