Abstractive Text Summarization Using Sequence-to-Sequence RNNs and Beyond

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. read more

PDF Abstract CONLL 2016 PDF CONLL 2016 Abstract

Datasets


Introduced in the Paper:

CNN/Daily Mail

Used in the Paper:

DUC 2004
Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Abstractive Text Summarization CNN / Daily Mail LEAD-3 ROUGE-1 40.42 # 28
ROUGE-2 17.62 # 30
ROUGE-L 36.67 # 30
Text Summarization CNN / Daily Mail (Anonymized) words-lvt2k-temp-att ROUGE-1 35.46 # 12
ROUGE-2 13.30 # 12
ROUGE-L 32.65 # 12
Text Summarization DUC 2004 Task 1 words-lvt5k-1sent ROUGE-1 28.61 # 10
ROUGE-2 9.42 # 9
ROUGE-L 25.24 # 8
Text Summarization GigaWord words-lvt5k-1sent ROUGE-1 36.4 # 22
ROUGE-2 17.7 # 21
ROUGE-L 33.71 # 24

Methods


No methods listed for this paper. Add relevant methods here