Rich Event Modeling for Script Event Prediction

Script is a kind of structured knowledge extracted from texts, which contains sequence events. Based on such knowledge, script event prediction aims to predict the subsequent event. To do so, two aspects should be considered for events, namely, description (i.e., what events contain) and encoding how they encoded). Most existing methods describe an by verb together with few core arguments subject, object, indirect object), are not precise enough. In addition, encoders limited fixed number arguments, flexible enough deal extra information. Thus, in this paper, we propose Rich Event Prediction (REP) framework prediction. Fundamentally, it based proposed rich description, enriches ones three kinds important information, senses verbs, semantic roles, types participants. REP extractor extract information texts. predictor then selects most probable The component transformer-based encoder that integrates above flexibly. Experimental results widely used Gigaword Corpus show effectiveness framework.