Contrastive Reasons Detection and Clustering from Online Polarized Debate

This work tackles the problem of unsupervised modeling and extraction of the main contrastive sentential reasons conv eyed by divergent viewpoints on polarized issues. It proposes a pipeline approach centered around the detection and clustering of phrases, assim ilated to argument facets using a novel Phrase Author Interaction Topic -Viewpoint model. The evaluation is based on the informativeness, the r elevance and the clustering accuracy of extracted reasons. The pipel ine approach shows a significant improvement over state-of-the-art meth ods in contrastive summarization on online debate datasets.