Multi−entity Sentiment Scoring

Abstract

We present a compositional framework for modelling entity-level sentiment (sub)contexts, and demonstrate how holistic multi-entity polarity scoring emerges as a by-product of compositional sentiment parsing. A data set of five annotators' multi-entity judgements is presented, and a human ceiling is established for the challenging new task. The accuracy of an initial implementation, which includes both supervised learning and heuristic distance-based scoring methods, is 5.6 6.8 points below the human ceiling amongst sentences and 8.1 8.7 points amongst phrases.