Acquiring a Dictionary of Emotion-Provoking Events

Abstract

This paper is concerned with the discovery and aggregation of events that provoke a particular emotion in the person who experiences them, or emotion-provoking events. We first describe the creation of a small manually-constructed dictionary of events through a survey of 30 subjects. Next, we describe first attempts at automatically acquiring and aggregating these events from web data, with a baseline from previous work and some simple extensions using seed expansion and clustering. Finally, we propose several evaluation measures for evaluating the automatically acquired events, and perform an evaluation of the effectiveness of automatic event extraction.