A RAG-BASED INTELLIGENT DOCUMENT ASSISTANT EMBEDDED WITHIN AN ERP-DRIVEN LEARNING MANAGEMENT SYSTEM
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
Students enrolled in enterprise-based resource planning (ERP) college portals routinely face a fundamental difficulty.The materials uploaded by their faculty members are accessible but not interactive.Browsing through multi-page PDF documents to find answers to specific questions is actually time-consuming and often results in an incomplete understanding.This study describes the design, construction, and field evaluation of an AI-driven document assistant embedded directly within an existing college ERP system built on Django.The assistant establishes a retrieval-augmented generation architecture in which faculty-uploaded PDF files are automatically interpreted, divided into overlapping text segments, converted into dense numerical representations by using a locally hosted sentence transformer model, and categorised inside a ChromaDB vector database under course-specific metadata labels.When a student types a question through the chat panel on the subject page, the system queries only the documents belonging to that subject, collects the five most contextually relevant segments, and passes them, along with the question, to the Google Gemini language model.The model then creates a response that draws exclusively from the retrieved course content, eliminating generic answers that are unrelated to the instructor's teaching style.Deployment at the Ramco Institute of Technology on the live course CCS369 -Text and Speech Analysis confirmed that the system consistently returns accurate, source-anchored answers, enforces strict subject-level data separation, and operates within the limits of a zero-cost embedding infrastructure and free-tier inference API.
Fields of Study & Subjects
Artificial Intelligence in Education
Topic Modeling
Information Retrieval and Search Behavior
Publication details
- Journal / source
- International Research Journal of Modernization in Engineering Technology and Science
- Publication date
- 28 March 2026
- Language
- EN