June 12, 2023

Research Assistant (Academic Student Employee)
Prosody modeling for emotion recognition
Application deadline: June 21, 2023 11:59 PM
Compensation: Per UW Policy, research assistants (RA) are paid at the student’s academic home department rate. The salary range provided is based upon a 50% FTE RA appointment.
The Department of Electrical & Computer Engineering at the University of Washington is a nationally ranked education and research program. Located on the main Seattle campus of the University of Washington, UW Electrical & Computer Engineering currently supports greater than 1100 undergraduate and graduate students and over 50 active research faculty. UW ECE is a dynamic organization that provides comprehensive administrative and advising services in support of our teaching, research and service missions.
Job Overview:
This project involves use of self-supervised methods to learn a general representation of prosody from unlabeled data, which is subsequently fine-tuned for emotion recognition. The work will build on existing neural network code for emotion recognition and for extracting prosodic features. The research will involve: i) exploring differences in usefulness of the approach for English vs. Chinese, and ii) integration of prosodic and text features. This project requires someone who is comfortable working with transformers and speech. 
  • Design experiments to explore the impact of different data sources and models on emotion recognition given speech audio. Modify existing code based on transformer models that combine self-supervised learning with fine-tuning on emotion-labeled data. Run experiments in multiple training and testing configurations.
  • Develop methods for integrating audio-only models of emotion with text-based models.
  • Contribute to monthly project reports, and document methods and results for a possible future paper.
  • Must be a current graduate student at the University of Washington eligible to work as a research assistant during the summer (registered at least 2 credits).
  • Familiarity with basic machine learning methods and experience running experiments using transformer-based neural network models. Experience working with speech or audio data is a plus.
  • Experience working with Python.