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

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 on. 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 [7] 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 can detect contradiction in online debates. To improve the classification results, more insight on coreference techniques and world knowledge that is hidden in the text should be extensively focused.