Project Name

RAG Based AI Chatbot for college

customer support
Project Team 15

Name - Arohan AR : TL22BTAI0222

Name - Adhithyan PC : TL22BTAI0416
Name - Muhammad Suhail : TL22BTAI0344
Name - Shahzad Bin Muhammad: TL22BTAI0343
 Dept Vision:
To evolve as a center of excellence for higher education in the field of Artificial Intelligence and
Machine Learning

Dept Mission:
M1: To equip the graduates with the latest technologies in the field of Artificial Intelligence and
Machine Learning
M2: To imbibe new knowledge by collaborating with industry and research organizations to mould
competent technocrats and entrepreneurs
M3: To instill societal and ethical responsibilities in all professional activities
INDEX

⮚ Motivation
⮚ Objective
⮚ Literature Review
⮚ Data Flow Diagram and Architecture
⮚ Implementation
⮚ Methodology
⮚ Challenges
⮚ Future developments
⮚ Conclusion
⮚ Reference
Motivation

• The driving motivation behind building an AI chatbot is to revolutionize the student experience
and streamline institutional efficiency.
• Key benefits of the system:
• Available 24/7
• Immediately addresses frequently asked questions
• Relieves human staff to focus on complex problems
• Expected outcomes:
• Improved student satisfaction through immediate practical support
• Meeting the expectations of digitally native students
• Reduced burden on overwhelmed support teams
• Lowered operating costs
• Improved total resource distribution in the college
 Objective
The primary objective of building a chatbot for college customer support is to provide efficient,
accessible, and immediate assistance to students, Parents and faculty regarding
common inquiries and support needs, ultimately enhancing the overall college experience
and streamlining operational processes.

Efficiency: Reduce the workload on human support staff

Accessibility: Offer 24/7 support

Immediacy: Deliver instant responses to common queries

Enhanced User Experience: Provide a convenient and user-friendly platform for accessing information and
support

Improved Operational Processes: Streamline support operations, reducing operational costs and improving
resource allocation within the college.

Increased Student Satisfaction: Ultimately, improve student satisfaction by providing readily available and
effective support services
Literature Review
Base Paper
A Chat Bot for Enrollment of Xi ’an Jiaotong-
Liverpool University Based on RAG*
Authors: Liwei.Xu,jiarui.Liu
Published in : IEEE,2024
• The paper "A Chat Bot for Enrollment of Xi’an Jiaotong-Liverpool University Based on RAG" is
relevant to the project as it explores AI-driven chatbot solutions for student query handling.

• Key features of the study:

• Uses Large Language Models (LLMs) for conversational AI.
• Incorporates Retrieval-Augmented Generation (RAG) for improved accuracy.
• Leverages university official data sources to ensure reliable responses.

• Benefits of using this approach in college customer support:

• Enhances query response accuracy by reducing AI hallucinations.
• Provides instant, reliable answers on admissions, academics, and university policies.
• Utilizes multi-source information retrieval to address diverse student inquiries.
• Supports continuous improvement with updated knowledge bases and filtering mechanisms.
Literature Review
Paper 1
RAG based Chatbot using LLMs
Authors: Dr. Vanishree K,Ananya G
Published in : IJSREM,2024
Research Focus – The study explores the use of Retrieval-Augmented Generation (RAG) and Large Language
Models (LLMs) in chatbot development.

Key Findings – AI-driven chatbots improve information retrieval by leveraging pre-trained models like GPT and
Llama2 for context-aware responses.

Relevance to College Support – Highlights the chatbot’s ability to provide immediate and accurate answers to student
and parent queries on admissions, academics, and administration.

Document-Based Knowledge Integration – Discusses methods to incorporate PDF-based information and FAQs,
aligning with the need for domain-specific chatbot capabilities.

Implications for Implementation – Suggests that RAG-based chatbots can enhance response accuracy, reduce
administrative workload, and improve student engagement.
Relevance of RAG-Based Chatbot for College Customer Support

1. RAG for FAQs – Efficiently retrieves and answers admission, academic, and student
queries.
2. PDF Processing – Extracts relevant info from college brochures, policies, and
handbooks.
3. LLM-Powered Responses – Uses Llama2, GPT-3.5 for accurate and contextual
replies.
4. Vector Database (FAISS) – Enables fast, semantic search for student and faculty needs.
5. User-Friendly Web Interface – Chainlit framework ensures easy access for users.
6. Cost Optimization – RL techniques minimize computational costs, making AI
affordable.
7. Scalable & Adaptive – Can expand to cover multiple departments & support queries.
Paper 2
Chatbots in customer service:Their relevance and impact on service quality,2022
Authors: Chiara Valentina Misischia, Flora Poecze, Christine Strauss
Published in:ELSEVIER

Enhancing Customer Interactions – AI chatbots improve response efficiency and optimize

business operations.
Core Technologies – Focus on natural language processing (NLP) and automation in
customer service.
Application in Higher Education – Relevant for student queries on admissions, course
registration, financial aid, and campus services.
Best Practices in Chatbot Design – Covers user engagement strategies and handling
complex queries effectively.
Implications for AI Support Systems – Insights help develop intelligent, student-centric
chatbot solutions for colleges.
Relevance of the Paper for AI Chatbot in College Support

1. AI & NLP – Uses AI & natural language processing for smart responses.
2. User-Friendly Interface – Ensures easy and engaging student interactions.
3. Automates Services – Handles admissions, course inquiries, and support.
4. System Integration – Connects with college databases for real-time info.
5. Context Awareness – Maintains conversation flow for better responses.
6. Data Security – Protects student data and ensures privacy.
7. Performance Metrics – Measures chatbot accuracy and user satisfaction.
Paper 3
College Enquiry Chatbot using Conversational AI
Authors:Dhruv Patel, Nihal Shetty, Paarth Kapasi, Ishaan Kangriwala
Published in: IJRASET,2023

Focus on College Chatbots – Explores chatbot development for college websites using Microsoft
Azure.

Comparative Analysis – Evaluates rule-based, Rasa NLU, and Azure-based chatbots for handling
student inquiries.

Key Technologies – Highlights NLP and machine learning for improving chatbot accuracy and user
experience.

Practical Implementation – Discusses chatbot integration with college websites and real-world
deployment strategies.

Insights for Our Project – Covers chatbot architecture, sentiment analysis, and active learning to
enhance response efficiency.
Relevance of the Paper for AI Chatbot in College Support

1. Conversational AI – Describe the use of LLM for human-like interactions.

2. Sentiment Analysis – Adapts responses based on user emotions.
3. Adaptive Learning – Improves continuously with user interactions.
4. 24/7 Support – Provides round-the-clock assistance for students & staff.
5. Cost Efficiency – Reduces admin workload and human intervention.
6. Multi-Platform Access – Works on web, mobile.
7. Data Insights – Analyzes student queries for service improvements.
Paper 4
Retrieval-Augmented Chatbots for Scalable Educational Support in Higher Education
Authors: Hassan Soliman, Hitesh Kotte, Miloš Kravčík, Norbert Pengel and Nghia Duong-Trung
Published in:2nd International Workshop on Generative AI for Learning Analytics At: Dublin, Ireland,2025

Focuses on Educational Support: Specifically addresses the development of RAG-based

chatbots for higher education, your exact project area.
Scalable and Personalized: Emphasizes scalable, personalized student support, directly
tackling a core challenge in large universities.
Combines LLMs with Learning Materials: Details the implementation of enriching LLM
based responses with pre-existing and contextual learning materials.
Discusses Technical Design: Focuses on the technological aspects of chatbot design and
LLM integration.
Evaluates Chatbot Performance: Assesses chatbot performance through human
evaluations and automated metrics, providing a framework for your own project's evaluation.
Retrieval-Augmented Chatbots for Scalable Educational Support in Higher
Education – Key Insights

1. Automates Enquiries – Reduces staff workload by handling student queries.

2. 24/7 Support – Provides round-the-clock assistance without human intervention.
3. NLP-Powered – Uses AI to understand and respond naturally to students.
4. Web Integration – Available on the college website for easy access.
5. AI Learning – Improves responses based on user interactions and feedback.
6. Multi-Platform Access – Works with messaging apps like Facebook Messenger.
Paper 5
Retrieval Augmented Generation (RAG) based Restaurant Chatbot with AI Testability
Authors:Vani Bhat, Sree Divya Cheerla, Jinu Rose Mathew, Nupur Pathak, Guannan Liu ,Jerry Gao
Published in: IEEE,2024

RAG for Relevant Answers: Shows RAG boosts chatbot accuracy, crucial for effective student
support.
Knowledge Graph Model: The approach aligns with your project's aim to organize and access
college resources.
Information Retrieval Success: Highlights the importance of retrieving right information to
ensure chatbot performance.
Contextual & Accurate Replies: Paper analyses models for contextual & accurate replies,
critical for students.
AI Testability Insights: Provides methods to evaluate and optimize chatbot performance and
measure key characteristics.
Personalized Support: Lessons from customer interactions are scalable for student-specific
chatbot.
Retrieval Augmented Generation (RAG) based Restaurant
Chatbot with AI Testability – Key Insights

1. AI & NLP-Powered – Uses AI, NLP, and ML for smart, human-like

interactions.
2. 24/7 Support – Always available to assist students anytime.
3. Cost-Efficient – Reduces staff workload and operational expenses.
4. Personalized Responses – Learns and adapts for better query handling.
5. Data Privacy & Security – Ensures student information is protected.
6. Proven Industry Use – The effectiveness of chatbot applications in
industry.
7. Future-Ready – Scalable for upgrades and AI advancements.
Data Flow Diagram
 Architecture
PDF data

Extract text Data User Query

Store it in vector Database as

Get Documents
chunks

Rank the Doc according to Context

Input The top 5 Docs and query into

LLM

Answer is generated according to

the Query and context provided
Current working Model
PDF uploader
 Main packages
• os – Handles file operations like deleting temporary files.

• tempfile – Creates temporary files for processing uploaded PDFs.

• chromadb – Implements a vector database for efficient document storage and retrieval.

• ollama – Calls the Ollama language model (LLM) to generate chatbot responses.

• streamlit – Provides a user-friendly web interface for the chatbot system.

• OllamaEmbeddingFunction (from chromadb.utils) – Converts text into vector embeddings for similarity search.

• PyMuPDFLoader (from langchain_community.document_loaders) – Extracts text from PDF documents.

• Document (from langchain_core.documents) – Defines structured text chunks for processing.

• RecursiveCharacterTextSplitter (from langchain_text_splitters) – Splits large text documents into smaller,

manageable chunks.

• CrossEncoder (from sentence_transformers) – Ranks retrieved documents based on relevance using a cross-
encoder model.

• UploadedFile (from streamlit.runtime.uploaded_file_manager) – Manages PDF uploads in the Streamlit app.

🔹 Key Technologies Used in the College Customer Support
Chatbot

1. Large Language Models (LLMs)

● Utilizes Gemma3:4B (a powerful AI model) for chatbot responses.

● Compared with other LLMs like LLaMA 2, tinyllama, and Llama3.2:3B for accuracy and integrity.

2. Retrieval-Augmented Generation (RAG)

● Improves response accuracy by fetching relevant documents before generating answers.

● Reduces AI hallucinations and ensures fact-based responses.

3. ChromaDB (Vector Database)

● Stores vector embeddings of documents for efficient retrieval.

● Uses cosine similarity to find the most relevant text.
🔹 Key Technologies Used in the College Customer Support
Chatbot

4. Cross-Encoder Model for Re-ranking

● Uses sentence-transformers CrossEncoder to improve document ranking.

● Ensures the most relevant information is prioritized in responses.

5. Ollama for AI Chatbot Responses

● Calls Llama3-based models to generate responses with structured prompts.

● Ensures chatbot follows guidelines for clarity and completeness.

6. Streamlit for User Interface

● Provides an interactive web-based chatbot.

● Supports file uploads (PDF, Excel) for document-based question answering.
Why RAG is Better for a College Customer Support Chatbot
1. Higher Accuracy

○ RAG reduces AI hallucinations by retrieving relevant, fact-based information from trusted sources like university databases and websites.
2. Context-Aware Answers

○ Retrieves relevant documents before generating a response, making the chatbot’s answers more contextually accurate and domain-specific.
3. Better Query Handling

○ Can process complex or uncommon questions by retrieving the most relevant documents instead of depending solely on pre-trained LLM
knowledge.
4. Multi-Source Integration

○ Allows integration of multiple data sources (FAQs, PDFs, websites, and academic policies) to improve response diversity and reliability.
5. Scalability & Efficiency

○ Optimizes large-scale document searching, making it faster and more efficient in retrieving precise information compared to traditional LLMs.
6. Improved Trust & User Satisfaction

○ Delivers more reliable and verifiable responses, increasing student confidence in the chatbot and reducing the need for human intervention.
Methodology

Methodology:
● Data Collection : Information is gathered from the web and publicly accessible
sources.
● Data Preprocessing: PDF data is processed and segmented to extract
meaningful text.
● Retrieval-Augmented Generation (RAG): Retrievers connect generative
models with external knowledge sources.
● User Interaction: A web application enables users to interact with the LLM.
● System Integration: The application integrates with both the database and the
generative model for seamless functionality.
Challenges Limitations
● Data Preprocessing : Need to change the ● Text Splitting: It sometimes causes
structure data was collected in so that context to be lost
when the text was split into chunks the ● Output Generation speed: The
context would not be lost
speed of output generated is slow
● Choosing the LLM: It was a tough choice
● Context Limit: There is a limit to the
when choosing the LLM. We needed to find
a good blance between LLM size, Context
amount of text that can be given to the
awarness, Accuracy and output speed. LLM as context.
● Choosing Chunk size: It was challenging ● The LLM may not full understand the
to find the appropriate size for each chunk query or may misunderstand the
of text to be divided. context
Future developments
➢ Add Voice assistance features

➢ Add multi-language assistance features

➢ Improve the division of text chunks stored in vector database

➢ Add a section for Topic explanations for each subject

 Conclusion

The AI chatbot for college customer support enhances student engagement

and streamlines administrative tasks through Retrieval-Augmented
Generation (RAG) and Large Language Models (LLMs). By providing
24/7 instant and accurate assistance, it reduces the workload on human
staff while improving accessibility for students, parents, and faculty.
With seamless integration into institutional systems, the chatbot efficiently
handles inquiries on admissions, financial aid, and campus services.. Future
enhancements may include Multilingual Support to assist a diverse student
body and Expanded Knowledge Base for integrating real-time updates on
college events
References

1.RAG based Chatbot using LLMs - RAG based Chatbot using LLMs PDF

2.Chatbots in customer service -

Chatbots in customer service:Their relevance and impact on service quality

3.College Enquiry Chatbot using Conversational AI -

College Enquiry Chatbot using Conversational AI

4.Chatbot for College Website - Chatbot for College Website

5.AI-Powered Chatbots for Customer Service -

AI-Powered Chatbots for Customer Service: Enhancing Customer Experience and Reducing C
osts