Self-Attention Guided Copy Mechanism for Abstractive Summarization

This paper incorporates TextRank into the attention map to amplify and polarize the attention scores for a better copy distribution. Given an attention map, a directed graph is built by treating the attention map as a soft adjacent matrix. By normalizing the adjacent matrix, a transition probability matrix is derived and iteratively refined by multiplying it multiple times. The final score matrix is added to the key states and normalized using a vanilla dot-product attention head. The generated copy distribution is matched with the original attention matrix to encourage the consistency with the attention mechanism by KL divergence.


  • TextRank is indeed an effective method in keyword extraction. It’s good to know it’s still working in neural models.
  • Idea is neat and it seems to have improved the ROUGE F1.
  • The KL loss is somewhat surprising, I can’t imagine that teaching an attention head to learn its amplified attention map actually worked. Is the KL loss necessary?
  • 5: Transformative: This paper is likely to change our field. It should be considered for a best paper award.
  • 4.5: Exciting: It changed my thinking on this topic. I would fight for it to be accepted.
  • 4: Strong: I learned a lot from it. I would like to see it accepted.
  • 3.5: Leaning positive: It can be accepted more or less in its current form. However, the work it describes is not particularly exciting and/or inspiring, so it will not be a big loss if people don’t see it in this conference.
  • 3: Ambivalent: It has merits (e.g., it reports state-of-the-art results, the idea is nice), but there are key weaknesses (e.g., I didn’t learn much from it, evaluation is not convincing, it describes incremental work). I believe it can significantly benefit from another round of revision, but I won’t object to accepting it if my co-reviewers are willing to champion it.
  • 2.5: Leaning negative: I am leaning towards rejection, but I can be persuaded if my co-reviewers think otherwise.
  • 2: Mediocre: I would rather not see it in the conference.
  • 1.5: Weak: I am pretty confident that it should be rejected.
  • 1: Poor: I would fight to have it rejected.

0 投票者