- Dr. Donald Brown (ESE) (Advisor)
- Dr. Laura Barnes (ESE) (Advisor)
- Dr. Michael Porter (ESE) (Chairperson)
- Dr. Nikolaos Sidiropoulos (ECE)
- Dr. Jennie L. Chiu (School of Education)
Zoom Meeting Information: Please email email@example.com for zoom information.
Title: Efficient collection, evaluation and deployment of large scale deep learning models in low resource natural language processing scenarios
Deep learning in natural language processing revolutionized low-resource domains like education and healthcare with approaches like transfer learning and prompting. Large language models can generalize to new tasks in low-resource domains; however, domain-specific data collection often beats generalized data for a given model size in terms of model performance.
My proposal aims to take the advances in deep learning and natural language processing and apply them in the context of low-resource domains like ``education''. I want to understand how we can efficiently train and deploy these large-scale models using approaches from tensor decomposition and improve the downstream application and deployment of such models.
Preliminary results show that tensor decomposition-based approaches can improve parameter efficiency over existing approaches by up to 5x. I plan to evaluate this approach's efficacy in education by deploying a real-world use case of a conversational agent system.