Inferring Affective Perceptions from Text using Natural Language Processing, Bayesian Inference and Affect Control Theory

• In Event: 099 - Section on Mathematical Sociology Paper Session. Open Topic on Mathematical Sociology

Abstract

Affect Control Theory (ACT) is a powerful tool in aiding our understanding of social events and the identities and behaviors within them. ACT scholars have developed both a mathematical model of social events and, through surveys, dictionaries detailing the affective meaning of many identities and behaviors. In a recently submitted article, we developed a novel methodology that utilized techniques from ACT, Natural Language Processing and Bayesian inference to infer perceptions of identities and behaviors in a large corpus of newspaper data relevant to the Arab Spring. Our efforts provide a unique approach to extracting affective meaning from text and a useful new tool for ACT scholars. In this proposal, we review this approach and results we obtained from it and then discuss three ways in which we plan to extend these previous efforts.

Authors

• Kenneth Joseph, Carnegie Mellon University

• Kathleen M. Carley, Carnegie Mellon University