An Adoption of a Contradiction Detection Task to Assist the Summarization of Online Debates

Abstract

In online debates, there are two opposing sides in which proponents and opponents sentimentally make arguments on various controversial topics. Currently, most debate summarization systems have focused on the generation of generic summaries. However, we view that these summaries may not entirely fulfill the needs of readers. On some occasions, readers may need to access the actual arguments that the proponents and opponents are debating. For these reasons, we aim to generate contradiction summaries from online debates. In this paper, we prepare new datasets grounded on the online debate summaries generated by Sanchan et al. (2017) and investigate whether a contradiction detection task could be effectively used to assist the generation of contradictory summaries for online debates. We observe which combination of features provides success in classifying contradiction. Our observation into the features and the qualitative analysis highlight that the employed features can detect contradiction in online debates. To improve the classification results, more insight into coreference techniques and world knowledge hidden in the text should be extensively focused on.

Keywords: online debate summarization, text summarization, contradiction detection, information extraction, sentence extraction

References

Download paper (PDF) Download BibTeX

Dataset: DEDX & DEDT datasets

@inproceedings{2020_Sanchan_Bontcheva_Aker,
  author    = {Sanchan, Nattapong and Bontcheva, Kalina and Aker, Ahmet},
  booktitle = {2020 - 5th International Conference on Information Technology (InCIT)},
  title     = {An Adoption of a Contradiction Detection Task to Assist the Summarization of Online Debates},
  year      = {2020},
  pages     = {185--190},
  month     = {Oct},
  doi       = {10.1109/InCIT50588.2020.9310941}
}

N. Sanchan, K. Bontcheva and A. Aker, “An Adoption of a Contradiction Detection Task to Assist the Summarization of Online Debates,” 2020 – 5th International Conference on Information Technology (InCIT), Chonburi, Thailand, 2020, pp. 185–190, doi: 10.1109/InCIT50588.2020.9310941.

More information

This paper is a continuing work in the Domain-Focused Summarization of Polarized Debates project:

1. Understanding Human Preferences for Summary Designs in Online Debates Domain
2. Gold Standard Online Debates Summaries and First Experiments Towards Automatic Summarization of Online Debate Data
3. Automatic Summarization of Online Debates
4. An Adoption of a Contradiction Detection Task to Assist the Summarization of Online Debates