Using Speech Emotions for Predicting the Check-Worthiness of Claims in Political Events

Document Type



Post-truth concerns today’s political scientists, who noticed that people tend to over- look factuality and consume information based on emotions. Grandiosity, informality, charisma, confidence, dominance, assertiveness, and emotions were analyzed as characteristics of public speakers’ tone of voice and were shown to affect people’s thoughts, attitudes, and favoritism regardless of the factuality of what the speaker is saying. In a way, the voice tone is an indirect way of communication that listeners infer and get influenced by. Modeling these vocal characteristics of public speakers is not straightforward. Among the many speech features studied in previous work, speech emotion extraction has been explored, which offers a more detailed vocal description. The focus is on extracting such speech emotional representation from political debates and speeches with the goal to advance the under-researched task of identifying check-worthy claims and examine the relationship be- tween emotions and check-worthy claims. The assumption is that speakers might choose a tone when they want to convey their most important messages. We propose two new ranking approaches to enhance check-worthiness detection. The first ranking approach demonstrated the usefulness of the ranking technique over classification for real-life scenarios for human fact-checkers. The second approach uncovered specific factors affecting the overall ranking performances. Specifically, emotion in descending order is more indicative of check-worthy claims. These results indicate that understanding the emotional change in the tone of the voice and contrasting it with the content of the speech is a promising direction for predicting the check-worthiness of claims in political debates and speeches. We believe that this framework can be further extended to other tasks related to factuality such as fact-checking, propaganda detection, framing, etc.

First Page


Last Page


Publication Date



Thesis submitted to the Deanship of Graduate and Postdoctoral Studies

In partial fulfillment of the requirements for the M.Sc degree in Natural Language Processing

Advisors: Dr. Shady Shehata, Dr. Preslav Nakov

Online access available for MBZUAI patrons