PhAITV: A Phrase Author Interaction Topic Viewpoint Model for the Summarization of Reasons Expressed by Polarized Stances
This work tackles the problem of unsupervised modeling and extraction of the main contrastive sentential reasons conveyed by divergent viewpoints in text documents. It proposes a pipeline framework that is centered around the detection and clustering of phrases, assimilated to argument facets using a novel Phrase Author Interaction Topic-Viewpoint (PhAITV) model. The evaluation is conducted on all the components of the framework. It is mainly based on the informativeness, the relevance and the clustering accuracy of extracted reasons. The framework shows a significant improvement over several configurations and state-of-the-art methods in contrastive summarization on online debate datasets.