Good News or Bad News: Using Affect Control Theory to Analyze Readers’ Reaction Towards News Articles

Human Language Technologies: The 2015 Annual Conference of the North American Chapter of the ACL, pages 1548–1558, Denver, Colorado, May 31 – June 5, 2015. c⃝2015 Association for Computational Linguistics Good News or Bad News: Using Affect Control Theory to Analyze Readers’ Reaction Towards News Articles Areej Ahothali David R. Cheriton School of Computer Science University of Waterloo Waterloo, Ontario, N2L3G1 aalhotha@cs.uwaterloo.ca Jesse Hoey David R. Cheriton School of Computer Science University of Waterloo Waterloo, Ontario, N2L3G1 jhoey@cs.uwaterloo.ca Abstract This paper proposes a novel approach to sen- timent analysis that leverages work in soci- ology on symbolic interactionism. The pro- posed approach uses Affect Control Theory (ACT) to analyze readers’ sentiment towards factual (objective) content and towards its en- tities (subject and object). ACT is a theory of affective reasoning that uses empirically de- rived equations to predict the sentiments and emotions that arise from events. This theory relies on several large lexicons of words with affective ratings in a three-dimensional space of evaluation, potency, and activity (EPA). The equations and lexicons of ACT were eval- uated on a newly collected news-headlines corpus. ACT lexicon was expanded using a la- bel propagation algorithm, resulting in 86,604 new words. The predicted emotions for each news headline was then computed using the augmented lexicon and ACT equations. The results had a precision of 82%, 79%, and 68% towards the event, the subject, and object, respectively. These results are significantly higher than those of standard sentiment analy- sis techniques. 1 Introduction Natural language texts are often meant to express or impact individuals’ emotions. Recognizing the underlying emotions expressed in or triggered by a text is essential to understanding the full meaning that a message conveys. Sentiment analysis (SA) researchers are increasingly interested in investigat- ing natural language processing techniques as well as emotion theories to identify sentiment expres- sions in natural language texts. They typically focus on analyzing subjective documents from the writer’s perspective using frequency-based word representa- tions and mapping them to categorical labels or sen- timent polarity, in which text (e.g., sentence or doc- ument) is associated, respectively, with either a de- scriptive label or a point in a continuum. In this paper, we focus on analyzing objective statements. Unlike subjective statements, that con- tain an explicit opinion or belief about an object or aspect (e.g., “Seventh Son is a terrible movie”), ob- jective statements present only factual information (e.g., “protestors were arrested in NYC”). Despite the limited research on fact-based sentiment analy- sis, many factual statements may carry or evoke sen- timents (e.g., news articles, blog comments. etc.). Thus, there is a need for a new approach that at- taches sentiment scores to objective (factual) state- ments as well as to the components that make them up. For example, a sentence like “x kills y” will clearly evoke a negative sentiment for the reader, and highly negative (yet different) sentiments towards each x and y (angry about x and sorry about y). Fur- ther, our affective evaluation of each entity …

Content truncated to 500 words