A Multi-Modal Evaluation of Chemistry Unbound using IRT and Sentiment Analysis
An award-winning NLP and Rasch IRT psychometric pipeline that evaluates student feedback and curriculum performance for university chemistry courses.
Machine Learning
Sentiment Analysis
Project Overview
Presentation Type: Poster Presentation (Emory Center for AI Learning Research Showcase).
Technologies Used: Python, Distilled-BERT, Rasch Item Response Theory (IRT) Analysis, NLP Pipeline.
Timeline: Jan 2025 – Present (Presented 2025).
My Role: Lead Researcher / AI Research Fellow.
Overview: Engineered an NLP and psychometric pipeline to evaluate the efficacy of Emory University's novel "Chemistry Unbound" curriculum. Combined Distilled-BERT sentiment classification with Rasch IRT modeling to analyze non-binary datasets of student assessment performance and feedback text.
Honors / Recognition: Best Poster Award & Emory AI X.perience Fellowship.
How to Access: Poster / Slides:
https://drive.google.com/file/d/1NW70DAgVd6Gc2rGh8FQfMCmYzYKHX0dg/view?usp=sharing