1362-GRM CogniTalk : A Grounded AI Learning Assistant for STEM Student: A Novel Multi-Stage Retrieval Approach
Location
https://campus.kennesaw.edu/colleges-departments/ccse/events/computing-showcase/fa26-cday-program.php
Document Type
Event
Start Date
20-11-2026 2:00 PM
Description
Students are often overwhelmed by course materials and turn to general AI tools that answer from the public internet rather than from actual course content producing answers that may not match what the professor taught. CogniTalk is an AI-powered educational assistant for students and lecturers that works in three stages: Knowledge Base Building uploads and processes course files into semantically coherent chunks stored in a vector database; Advanced RAG rewrites queries, runs dual vector searches, and applies LLM reranking to fetch the most relevant content; and Generation delivers six grounded features Q&A, practice tests, concept summaries, Socratic dialogue, visual study maps, and voice I/O. Evaluated on 60 questions with 14 students, CogniTalk improved accuracy by 30% and completeness by 29%; 79% would recommend it.
1362-GRM CogniTalk : A Grounded AI Learning Assistant for STEM Student: A Novel Multi-Stage Retrieval Approach
https://campus.kennesaw.edu/colleges-departments/ccse/events/computing-showcase/fa26-cday-program.php
Students are often overwhelmed by course materials and turn to general AI tools that answer from the public internet rather than from actual course content producing answers that may not match what the professor taught. CogniTalk is an AI-powered educational assistant for students and lecturers that works in three stages: Knowledge Base Building uploads and processes course files into semantically coherent chunks stored in a vector database; Advanced RAG rewrites queries, runs dual vector searches, and applies LLM reranking to fetch the most relevant content; and Generation delivers six grounded features Q&A, practice tests, concept summaries, Socratic dialogue, visual study maps, and voice I/O. Evaluated on 60 questions with 14 students, CogniTalk improved accuracy by 30% and completeness by 29%; 79% would recommend it.