An Event-local View: Emotion Interplay in the Underlying Social Graph of a Literary Text

Abstract

The traditional view of a social graph is static, though the graph itself evolves with events. An event-local view can capture this dynamic evolution and offer an event-specific local view of the social graph. In this paper, a novel framework is proposed to offer such a view of the underlying social network of a literary text accentuated with the emotion inter-play. It does this by integrating several pieces together i.e. extraction of character names with co-reference resolution, sentiment based windowing to identify main events in the text, detection of the underlying social graph of characters based on co-occurrence in the context of an event and supervised learning to detect emotions between characters. It delivers an integrated visualization to display the emotion interplay between main protagonists in the underlying social network in an event-specific manner. This is in contrast to the traditional static global views of the social graph made by analyzing the full text. Our work, by adopting an event-local view successfully captures the dynamic and evolving aspect of the underlying social graph.