Abstractive Text Summarization Using Sequence-to-Sequence RNNs and Beyond (original) (raw)

View PDF

Abstract:In this work, we model abstractive text summarization using Attentional Encoder-Decoder Recurrent Neural Networks, and show that they achieve state-of-the-art performance on two different corpora. We propose several novel models that address critical problems in summarization that are not adequately modeled by the basic architecture, such as modeling key-words, capturing the hierarchy of sentence-to-word structure, and emitting words that are rare or unseen at training time. Our work shows that many of our proposed models contribute to further improvement in performance. We also propose a new dataset consisting of multi-sentence summaries, and establish performance benchmarks for further research.

Submission history

From: Ramesh Nallapati [view email]
[v1] Fri, 19 Feb 2016 02:04:18 UTC (34 KB)
[v2] Mon, 11 Apr 2016 22:50:03 UTC (163 KB)
[v3] Sat, 23 Apr 2016 02:38:01 UTC (163 KB)
[v4] Wed, 10 Aug 2016 22:56:10 UTC (314 KB)
[v5] Fri, 26 Aug 2016 16:13:13 UTC (314 KB)