Words of Persuasion – The Text Predictors of Persuasive TED Talks

What makes for a persuasive presentation? How can you speak to persuade? To create theodorespeaks.com, I webscraped every TED Talk, analyzed words in the transcripts & user’s ratings, and found that the most persuasive speakers used MORE negative emotion words and LESS ‘I, me, my’ words.

So the next time you’re trying to convince, try calling attention to the negative emotions (‘sadness, frustration, anger’) the problem causes and to the ways ‘we,’ ‘us,’ or ‘you’ — not ‘I’ can solve them. At theodorespeaks.com, you can also paste in the text of your own speech to see if it’s likely to be persuasive or not…

TheodoreSpeaks.com is the result of a natural language processing project to reveal linguistic and psychological features that predict a persuasive TED Talk. I webscraped 2600+ TED Talk transcript and it’s metadata through 2017 and then used decision trees, random forest regressors, and linear regression to find key predictors of persuasive ratings by viewers.

I found that the change in negative and positive emotion words across the talk and the speaker’s use of key social pronouns like “I” and “we” made a big impact on persuasive ratings.

My analyses resulted in important categories of words that make up a “linguistic signature” of persuasion and a classifier that you can use to predict the persuasiveness of your own text.

For professionals who need to communicate and influence others – TheodoreSpeaks.com is a data product that uses natural language processing techniques and statistical modeling to provide insights on how to speak to persuade.