Date of Degree

9-2018

Document Type

Dissertation

Degree Name

Ph.D.

Program

Computer Science

Advisor

Rivka Levitan

Committee Members

Michael Mandel

Alla Rozovskaya

Andrew Rosenberg

Subject Categories

Artificial Intelligence and Robotics

Keywords

self-reported personality, deep learning, multi-task learning

Abstract

Personality aims at capturing stable individual characteristics, typically measurable in quantitative terms, that explain and predict observable behavioral differences. Personality has been proved to be very useful in many life outcomes, and there has been huge interests on predicting personality automatically. Previously, there are tremendous amount of approaches successfully predicting personality. However, most previous research on personality detection has used personality scores assigned by annotators based solely on the text or audio clip, and found that predicting self-reported personality is a much more difficult task than predicting observer-report personality. In our study, we will demonstrate how to accurately detect self-reported personality from speech using various technique include feature engineering and machine learning algorithms. Individual speaker differences such as personality play an important role in deception detection, adding considerably to its difficulty. We therefore hypothesize that personality scores may provide useful information to a deception classifier, helping to account for interpersonal differences in verbal and deceptive behavior. In final step of this study, we focus upon the personality differences between deceivers as well as their common characteristics. We helped collect within- and cross-cultural data to train new automatic procedures to identify deceptive behavior in American and Mandarin speakers. We examined whether personality recognition can help to predict individual differences in deceivers’ behavior. Therefore, we embedded personality recognition classifier into the deception classifier using deep neural network to improve the performance of deception detection.

This work is embargoed and will be available for download on Tuesday, September 10, 2019

Graduate Center users:
To read this work, log in to your GC ILL account and place a thesis request.

Non-GC Users:
See the GC’s lending policies to learn more.

Share

COinS