A Repository of Machine Summaries for Generic News Summarization on DUC 2004

This repository includes summaries produced by several state-of-the-art summarization systems and popular baseline systems on DUC 2004 task 2.

We provide summaries from the following systems:
    - Baseline systems
          FreqSum (probability), TsSum (topic signatures), Centroid, Cont. LexRank, GreedyKL
    - State-of-the-art systems
          CLASSY 04 (Peer 65), CLASSY 11, DPP, ICSISumm, OCCAMS_V, RegSum, Submodular

The summaries are available here: [link]
A layout of the corpus can be found in the README file.

More details about implementation of the systems, choices of ROUGE settings, pairwise comparison between systems and summary overlap at different levels are in our paper:

Kai Hong, John M. Conroy, Benoit Favre, Alex Kulesza, Hui Lin, and Ani Nenkova
A Repository of State of the Art and Competitive Baseline Summaries for Generic News Summarization
In Proceedings of LREC, 2014 [pdf]

Related Links