Transforming course feedback: personalized opine surveys using language models

Abstract

This B.Tech project at IIIT-Delhi reimagines the traditional course feedback system, ”opine,” by transitioning to a personalized model using Language Models (LLMs). The project is designed for flexibility, allowing easy integration of various LLMs, including potential open-source models. Initial implementation focuses on the Operating Systems course, employing GPT- 3.5 through OpenAI’s API endpoints and the Semantic Kernel SDK. User engagement involved selecting 20 students for the opine survey, with qualitative feedback which will be gathered through interviews. Technical implementation integrates GPT-3.5, with preliminary feedback providing valuable insights. Future work includes scalability solutions using embeddings, an extensive response sheet based on vector embeddings, and dynamic survey customization for instructors. This project aims to optimize course feedback, providing a foundation for dynamic, personalized surveys applicable across diverse courses. The report outlines methodology, results, and future directions for continued improvement, contributing to an enhanced feedback process at IIIT-Delhi.