Automatic Detection of Verbal Deception (Synthesis Lectures on Human Language Technologies)

A growing field in computer applications is the use of algorithms to spot the lie. The most promising area within this field is the analysis of the language of the liar since speakers effectively control only the meaning they wish to convey, but not the linguistic style of the communication. With the advent of computational means to analyze language, we now have the ability to recognize differences in the way speakers phrase their lies as opposed to their truths.

The main goal of this book is to cover the advances of the last 10 years in automatically discriminating truths from lies. To give the reader a grounding in deception studies, it describes a range of behaviors (physiological, gestural as well as verbal) that have been proposed as indicators of deception. An overview of the primary psychological and cognitive theories that have been offered as explanations of deceptive behaviors gives context for the description of specific behaviors. The book also addresses the differences between data collected in a laboratory and real-world data with respect to the emotional and cognitive state of the liar.

It discusses sources of real-world data and problematic issues in its collection and identifies the primary areas in which applied studies based on real-world data are critical, including police, security, border crossing, customs, and asylum interviews; congressional hearings; financial reporting; legal depositions; human resource evaluation; predatory communications that include Internet scams, identity theft, fraud, and false product reviews. Having established the background, the book concentrates on computational analyses of deceptive verbal behavior, which have enabled the field of deception studies to move from individual cues to overall differences in behavior. The book concludes with a set of open questions that the computational work has generated.