SURF Mentoring
Potential projects/topics: Students have the opportunity to select from the following research projects:
- Project 1: This project focuses on developing and evaluating methods for identifying and classifying vocalizations or behaviors from raw wearable sensor data from animals (wild baboons, goats, sheep). The student will compare different classification approaches using an existing dataset. Emphasis will be on assessing the performance and tradeoffs of different classification approaches, but students may also use processed data to answer behavioral questions, if interested
- Project 2: This project focuses on applying an existing computational model of conversational dynamics to conversation data to evaluate the underlying social dynamics. The student may work with either publicly available annotated datasets (e.g., television shows, podcasts) or synthetically generated conversations (LLMs) to examine how the model behaves under different data conditions. The project will focus on evaluating how model outputs are shaped by conversational characteristics.
- Project 3: This project focuses on developing and evaluating methods for representing the semantic content of individual speaking turns within a conversation. Using an existing dataset collected from student teams, the student will use different language models to embed conversational statements in semantic space to evaluate how well they capture the underlying intent and opinions of speakers.
Potential skills gained: data analysis, machine learning, computational modeling, natural language processing, data visualization, programming in Python or R
Required qualifications: Some programming experience in Python or R is preferable. Best suited to majors in ECE, Biosciences, or Cognitive Sciences
Direct mentor: Faculty/P.I.
