Authors: Joel Young and Pedro Ortiz
We answer the question: “Can we learn to identify persuasion as characterized by Cialdini’s persuasion model using traditional machine learning techniques?” Leveraging three machine learning tools: naive bayes, maximum entropy, and support vector machines combined with gappy and orthogonal sparse bigrams weare able to develop several weak learners with F-Scores significantly better than random. The research is based onthe NPS Persuasion Corpus consisting of 37 transcripts from four hostage negotiation situations. Each utterance in the corpus was hand annotated for one of nine categories of persuasion based on Cialdini’s model: reciprocity, commitment,consistency, liking, authority, social proof, scarcity, other, and not persuasive. Ability to automatically detect persuasion is essential for machine interaction on the social web.
Keywords: persuasion detection, machine learning, semanticcomputing, artificial intelligence