Wojciech Kryściński, Romain Paulus, Caiming Xiong, Richard Socher
Abstractive text summarization aims to shorten long text documents into a human readable form that contains the most important facts from the original document. However, the level of actual abstraction as measured by novel phrases that do not appear in the source document remains low in existing approaches. We propose two techniques to improve the level of abstraction of generated summaries. First, we decompose the decoder into a contextual network that retrieves relevant parts of the source document, and a pretrained language model that incorporates prior knowledge about language generation. Second, we propose a novelty metric that is optimized directly through policy learning to encourage the generation of novel phrases. Our model achieves results comparable to state-of-the-art models, as determined by ROUGE scores and human evaluations, while achieving a significantly higher level of abstraction as measured by n-gram overlap with the source document.
| Task | Dataset | Metric | Value | Model |
|---|---|---|---|---|
| Text Summarization | CNN / Daily Mail (Anonymized) | ROUGE-1 | 40.02 | ML+RL ROUGE+Novel, with LM |
| Text Summarization | CNN / Daily Mail (Anonymized) | ROUGE-2 | 15.53 | ML+RL ROUGE+Novel, with LM |
| Text Summarization | CNN / Daily Mail (Anonymized) | ROUGE-L | 37.44 | ML+RL ROUGE+Novel, with LM |
| Text Summarization | CNN / Daily Mail | ROUGE-1 | 40.19 | ML+RL ROUGE+Novel, with LM |
| Text Summarization | CNN / Daily Mail | ROUGE-2 | 17.38 | ML+RL ROUGE+Novel, with LM |
| Text Summarization | CNN / Daily Mail | ROUGE-L | 37.52 | ML+RL ROUGE+Novel, with LM |
| Abstractive Text Summarization | CNN / Daily Mail | ROUGE-1 | 40.19 | ML+RL ROUGE+Novel, with LM |
| Abstractive Text Summarization | CNN / Daily Mail | ROUGE-2 | 17.38 | ML+RL ROUGE+Novel, with LM |
| Abstractive Text Summarization | CNN / Daily Mail | ROUGE-L | 37.52 | ML+RL ROUGE+Novel, with LM |