Speaker:
Dr. Joan Bachenko
Linguistech Consortium
Deception Detection Technologies

Time: Wednesday 05/06/2009 3-4PM
Location: Babbio 110

Abstract:

This talk describes a NLP approach to the identification of deceptive language in transcribed interviews. Although deception detection has long been an interest among research psychologists and law enforcement professionals, language-based analyses of deception is a new area of investigation for computational linguistics.

The talk will have three parts: a brief review of previous work on language and deception detection that has motivated our research, a description of the model we have developed for identifying deceptive language by native speakers and a review of an experiment that tested the ability of the model to identify deceptive and non-deceptive passages of transcribed speech. The data used in our work comes exclusively from "real world" sources--police interrogations, criminal statements and legal depositions. Using Classification and Regression Tree techniques, we found that the model correctly identifies 74.9% of the deceptive and non-deceptive propositions rated in the experiment. Current applications of the model focus on automatic recognition of deceptive narratives in large corpora and aids to real-time interviewing.

Biography:

Dr. Bachenko received her Ph.D. in Linguistics from New York University. Her work in computational linguistics has focused on research and technology development in natural language processing, speech synthesis and speech recognition. She has spent roughly half her career at research laboratories—the Naval Research Laboratory in Washington, D.C. and Bell Laboratories in Murray Hill, NJ. She left Bell Laboratories to co-found Linguistic Technologies, Inc. (LTI), a Minnesota-based startup that developed speech recognition technology for the transcription of medical dictation over the phone. After LTI was acquired, Dr. Bachenko began her ongoing collaboration with Montclair State University in New Jersey. Her research with Montclair focuses on the analysis of deceptive language in native and non-native speakers of English. She is currently working on the implementation of a NLP model of deceptive language and on the development of training methods that will enable interviewers to detect significant language changes in real time. Dr. Bachenko has also served as adjunct faculty at the University of Minnesota and Montclair State University.

For more information please contact:

Yingying Chen
Assistant Professor & NIS Graduate Program Director
Burchard
Room 210
Phone: 201.216.8066
Fax: 201.216.8246
yingying.chen@stevens.edu

seminar_05062009