Academic Assistant :
Date
1-Feb-2025
Journal Title
Journal ISSN
Volume Title
Publisher
Comilla University
Abstract
The Academic Assistant is an AI-powered robot designed to enhance educational
and research experiences. It leverages a fine-tuned DeepSeek 1.5B language model
with Retrieval-Augmented Generation (RAG) to provide accurate, context-aware
responses to academic queries. The system integrates a humanoid robot with
sensors, a Raspberry Pi 3, and Google TTS for interactive, real-time engagement.
Key metrics show a BLEU score of 0.480, ROUGE-1 of 0.820, ROUGE-2 of 0.700,
and ROUGE-L of 0.820, indicating high accuracy and recall. By continuously
learning from user feedback through reinforcement learning, it improves response
quality. This aims to revolutionize education by offering personalized, reliable
academic support, making learning more accessible and engaging for students,
educators and researchers alike
and research experiences. It leverages a fine-tuned DeepSeek 1.5B language model
with Retrieval-Augmented Generation (RAG) to provide accurate, context-aware
responses to academic queries. The system integrates a humanoid robot with
sensors, a Raspberry Pi 3, and Google TTS for interactive, real-time engagement.
Key metrics show a BLEU score of 0.480, ROUGE-1 of 0.820, ROUGE-2 of 0.700,
and ROUGE-L of 0.820, indicating high accuracy and recall. By continuously
learning from user feedback through reinforcement learning, it improves response
quality. This aims to revolutionize education by offering personalized, reliable
academic support, making learning more accessible and engaging for students,
educators and researchers alike
Description
Keywords
L - Location, O - Objective, C - Constraints, Technological Constraints, Performance Constraints, Learning Constraint, S - Stakeholders, Students, Educators, Researchers, H - Human Factors