Fostering User Engagement: Rhetorical Devices for Applause Generation Learnt from TED Talks

One problem that every presenter faces when delivering a public discourse is how to hold the listeners’ attentions or to keep them involved. Therefore, many studies in conversation analysis work on this issue and suggest qualitatively constructions that can effectively lead to audience’s applause. To investigate these proposals quantitatively, in this study we analyze the transcripts of 2,135 TED Talks, with a particular focus on the rhetorical devices that are used by the presenters for applause elicitation. Through conducting regression analysis, we identify and interpret 24 rhetorical devices as triggers of audience applauding. We further build models that can recognize applause-evoking sentences and conclude this work with potential implications. Introduction and Motivation Many academic studies have been devoted to the tasks of user engagement characterization. However, applause, as the most direct audience reaction, has till now not been fully investigated and understood. The audience do not just clap whenever they like, they do so only at certain points and are continuously looking for appropriate times when applause can possibly occur (Kuo, 2001). Having a deep understanding of audience’s applause is important because it will help the speakers to better design their speech with recognizable calls for conjoined response, and to make their presentation more appealing and engageable. Despite its importance, to date relatively limited work have been conducted on the topic of applause generation, except a few qualitative studies done by social psychologists. Atkinson (1984) first claimed that applause is closely synchronized with a number of actions on the part of the speakers, which he referred to as “rhetorical devices”. He identified three rhetorical devices that are effective in evoking applauses, including: contrasts, three-part lists, and projection of names. Heritage and Greatbatch (1986) found five other basic rhetorical devices, including: puzzlesolution, headline-punchline, combination, position taking This is a pre-print of an article appearing at ICWSM 2017. and pursuit. In addition, new categories were also identified in many recent studies, such as greetings, expressing appreciations, etc. (Bull and Feldman, 2011). To date, research on applause generation has been limited to the analysis of political speeches only. Besides, all of the aforementioned work were conducted using qualitative methods. Many critical questions, such as, What triggers the audience’s applause?, When do audience applaud?, etc., remain unanswered. To address these gaps, this work aims to identify the rhetorical devices for applause elicitation using data-driven methods. To this end, we propose two research questions: RQ1: What are the rhetorical devices that cause the audiences to applaud during a specific part of a speech? RQ2: To what extent the hypothesized rhetorical devices can be used to predict applause generation? To answer both questions, we crawl 2,135 TED talk transcripts and conduct quantitative analysis to investigate the factors that could trigger audience’s applause. We find that factors such as, gratitude expressions, phonetic structure, projection of names, emotion, etc., have significant effects on applause generation. These identified factors are later used to build machine learning models that can automatically identify applause-evoking segments.