Research &  Lab Work:

Automatic Sarcasm Detection in Speech

As a part of my Master’s thesis, I built a system for detecting sarcasm in speech.  This system uses a baseline of acoustic features that have been found to be helpful in human sarcasm identification. The system also uses an effective way of modeling and applying prosodic contours to the task of automatic sarcasm detection.  This approach applies sequential modeling to categorical representations of pitch and intensity contours obtained via k-means clustering. Using a SimpleLogistic classifier, we are able to predict sarcasm with 81.57% accuracy. This result suggests that certain pitch and intensity contours are predictive of sarcastic speech.

Future work on this project to come.

Prosody Transfer from L1 Mandarin to L2 English.

Abstract to be posted sometime in June 2014.

Prosody Modeling for Non-Native Speech.

Pending!  Stay tuned.


Set your Twitter account name in your settings to use the TwitterBar Section.