Extreme Summarization

Description

This is a demo of Topic-Aware Convolutional Neural Networks for Extreme Summarization.

We compare outputs of three different systems: LEAD (first sentence in the article), ConvS2S (convolutional sequence -to-sequence model) and Topic-ConvS2S (topic-aware convolutional sequence-to-sequence model).

Our newly collected dataset and the source code for our generator can be found on github.

See here for current results on the XSum dataset we created.

Paper

To read more about XSum, see the paper that can be downloaded here.

@InProceedings{D18-1206,
   author =     "Narayan, Shashi and Cohen, Shay B. and Lapata, Mirella",
   title =      "Don't Give Me the Details, Just the Summary! Topic-Aware Convolutional Neural Networks for Extreme Summarization",
   booktitle =  "Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing",
   year =       "2018",
   publisher =  "Association for Computational Linguistics",
   pages =      "1797--1807",
   location =   "Brussels, Belgium",
   url =        "http://aclweb.org/anthology/D18-1206"
}

Demo


Text:

LEAD Summary Baseline (first sentence in the article)

ConvS2S

Topic-ConvS2S

LOG

We thank Nikos Papasarantopoulos and Shashi Narayan for helping with the development of this demo.