“Good Luck with That!”: Teaching Machines to Detect Sarcasm

Authors

Xinyi Ou, Grand Valley State University

Keywords

sarcasm detection, machine learning, audio signal processing, text-based analysis

Disciplines

Computer Sciences

Mentor

Gregory Wolffe

Abstract

Sarcastic text messages run the risk of misinterpretation, primarily because the reader does not have access to the sender’s body language or tone of voice to help intuit the original intentions of the author. However, if machines could be taught to accurately detect sarcasm, those on the receiving end could be forewarned about the presence of sarcastic content. This is more than a frivolous pursuit; similar software is employed by the U.S. government to distinguish serious threats against the country from sarcastic commentary. While current sarcasm detection is based on word analysis, the increasing popularity of dictation (speech-to-text) software has made additional vocal cues available for machine use. Using a collection of sarcastic/sincere audio clips of the main character Lorelei Gilmore from Gilmore Girls, this project combined machine-based text analysis with voice/tone signal processing analysis to build a more complex—and hopefully more accurate—model of sarcasm detection.

ScholarWorks Citation

Ou, Xinyi, "“Good Luck with That!”: Teaching Machines to Detect Sarcasm" (2015). Honors Projects. 450.
https://scholarworks.gvsu.edu/honorsprojects/450