Department of Computer Engineering
Study Partner – Your AI Learning
Companion
Team Member 1 - Minav Karia
Team Member 2 - Yashank Kothari
Team Member 3 - Aarush Saboo
Outline of Presentation
□ Motivation
□ Abstract
□ Problem Statement and Objective
□ Literature Survey
□ Implementation plan
□ Flow diagram
□ Conclusion
□ References
Motivation
To improve education by offering accessible video learning, interactive quizzes,
and instant AI chatbot support creating an engaging, personalized, and effective
self-paced learning experience for everyone, anywhere, anytime.
Abstract
● AI-powered academic assistant using the 4-Quadrant MOOC model
● Delivers personalized content: video, text, quizzes, chatbot
● Supports all branches and semesters, aligned with syllabus
● Real-time query resolution via NLP-based chatbot
● Multilingual, accessible, and mobile-friendly platform
Problem Statement and Objective
Students lack a smart, unified platform that supports personalized academic learning.
Existing MOOCs are fragmented, not syllabus-aligned, and don’t offer instant help.
There's a need for an AI-based solution that connects video, text, quizzes, and chatbot
support.
Objectives
● Develop an AI-powered platform supporting all 4 MOOC quadrants for holistic
academic support.
● Provide personalized learning based on course, branch, and semester.
● Enable real-time query resolution through AI chatbot and discussion forums.
● Centralize access to video lectures and textual content across all subjects.
● Enhance student performance via targeted quizzes and self-assessments.
● Deliver inclusive and multilingual educational content for diverse learners.
● Integrate YouTube and academic videos for engaging, interactive learning.
● Track learner progress and adapt recommendations accordingly.
Literature Survey
PUBLICATION YEAR TITLE OF THE PAPER OBJECTIVE GAPS IDENTIFIED
To classify user intent in Limited to
dialogue systems using deep single-turn intents;
Intent Classification with BERT, ACL – Zhang lacks academic
contextual models.
et al.
domain
specialization.
2020
No
To generate natural questions
personalization or
from input texts using
syllabus
Harvesting Paragraph-Level Q&A Pairs, ACL – attention-based seq2seq
adaptation;
Du & Cardie models.
focused on
2018 general datasets.
Literature Survey
PUBLICATION YEAR TITLE OF THE PAPER OBJECTIVE GAPS IDENTIFIED
Review architectures and Most chatbots lack
use-cases of chatbots in real-time NLP
A Survey on Educational Chatbots, IEEE understanding and
digital learning.
Access – Ruan et al.
syllabus linkage.
2022
Does not support
multilingual
Suggest content to students
personalization or
Personalized MOOC Recommendation with using NLP-driven learner
quiz integration.
NLP, JEDM – Sweeney et al. profiles.
2021
Literature Survey
PUBLICATION YEAR TITLE OF THE PAPER OBJECTIVE GAPS IDENTIFIED
Generate subject-specific Ignores learner
quizzes using pre-trained performance
Adaptive Quiz Generation using Transformer
transformers. feedback in
Models, IEEE TLT – Kim et al.
generation loop.
2023
Focuses on
translation but
Explore NLP-driven content
lacks integration
Multilingual NLP for Inclusive Education, ACM translation and accessibility
into complete
TALLIP – Rajan et al. across Indian languages.
learning
2024 pipelines.
Implementation plan
Tech Stack
□ Frontend: ReactJS, Tailwind CSS, Electron js
□ Backend: [Link] + Express, Tensorflow
□ Database: MongoDB, Supabase, Pinecone
□ AI/NLP Models: Gemini API, T5, BERT, Sentence-BERT
□ Tools: Hugging Face Transformers, Faiss (semantic search),
LangChain
□ Deployment: Vercel (frontend), Docker (containers), Google
Colab (model tuning), Pythonanywhere
Implementation plan
Dataset
● NPTEL Video Captions
○ Extracted using YouTube's transcript feature or
youtube-transcript-api
○ NPTEL Portal: [Link]
○ YouTube Channel: [Link]
● KJ Somaiya School of Engineering Syllabus
● MCQ & Quiz Bank Datasets
○ Examveda – Subject-wise MCQs
○ MCQDB – Publicly shared question sets
○ Kaggle Quiz Datasets: [Link]
Implementation plan
Flow of System
● Student enters query or selects topic
● Gemini API processes query intent
● Backend fetches relevant text/video content + generates quiz
● Chatbot returns output + recommends next steps
● User progress tracked for adaptive content delivery
Diagram Elements:
User → [Chatbot] → [NLP Layer] → [Content Fetch + Quiz Gen] → [UI Output + Analytics]
Flow Diagram
Conclusion
● An AI-powered learning platform mapped to the MOOC 4-Quadrant model.
● Uses NLP and Gemini API to enable personalized, syllabus-aware education.
● Integrates text, video, quizzes, and chatbot into one cohesive system.
● Multilingual, inclusive, and performance-driven.
● Scalable for all branches, years, and learner types.
References
1. Zhang, C., Wu, Y., & Yu, P. S. (2020). Intent Classification with BERT. Proceedings of ACL 2020.
[Link]
2. Du, X., & Cardie, C. (2018). Harvesting Paragraph-Level Question-Answer Pairs. Proceedings of ACL 2018.
[Link]
3. Sweeney, M., Lester, J., & Mott, B. (2021). NLP-Based Recommendation for MOOCs. Journal of Educational Data Mining (JEDM).
[Link]
4. Google Developers. Gemini API Documentation.
[Link]
5. Hugging Face. Transformers Library.
[Link]
6. National Programme on Technology Enhanced Learning (NPTEL).
[Link]
7. University Grants Commission (UGC) – Syllabus Portal.
[Link]
8. Central Board of Secondary Education (CBSE) Academic Curriculum.
[Link]
9. ExamVeda – MCQ Repository.
[Link]
10. Kaggle – Public Educational Datasets.
[Link]