Leading University

Department of Computer Science and Engineering

CSE 4801

Eduverse
Koushik Roy 2012020173
Prokash Paul Pollob 2012020390

Department of Computer Science and Engineering

Supervisor
Prithwiraj Bhattacharjee
Lecturer
Department of Computer Science and Engineering

7th March, 2024

 Eduverse

A Project submitted to the

Department of Computer Science and Engineering
Leading University
Sylhet - 3112, Bangladesh
in partial fulfillment of the requirements for the degree of
Bachelor of Science in Computer Science and Engineering

By
Koushik Roy 2012020173
Prokash Paul Pollob 2012020390

Supervisor
Prithwiraj Bhattacharjee
Lecturer
Department of Computer Science and Engineering

7th March, 2024

 Recommendation Letter
The project entitled Eduverse submitted by the students

1. Koushik Roy

2. Prokash Paul Pollob

is a record of research work carried out under my supervision and I, hereby, approve that the report
be submitted in partial fulfillment of the requirements for the award of their Bachelor Degrees.

Signature of the Supervisor:

Prithwiraj Bhattacharjee
Date: 7th March, 2024
 Certificate of Acceptance of the Project
The project entitled Eduverse
submitted by the students

1. Koushik Roy

2. Prokash Paul Pollob

on 7th March, 2024

is, hereby, accepted as the partial fulfillment of the requirements for the award of their Bachelor
Degrees.

Head of the Dept. Convener, Project Committee Supervisor

Rumel M. S. Rahman Pir Arafat Habib Quarishi Prithwiraj Bhattacharjee
Assistant Professor & Head Advisor Lecturer
Department of Computer Department of Computer Department of Computer

Science and Engineering Science and Engineering Science and Engineering

 Abstract

Our project proposes an integrated classroom management system featuring PDF and YouTube
chatbots, along with assignment management capabilities. The PDF chatbot facilitates resource ac-
cess through natural language processing, enabling students to retrieve course materials and lecture
notes seamlessly. Additionally, the YouTube chatbot enhances multimedia engagement by linking
educational videos and fostering interactive discussions. The assignment management module
allows teachers to post assignments and deadlines while enabling students to submit assignments
electronically and track their progress. This centralized platform streamlines communication,
promotes self-directed learning, and facilitates timely feedback. By leveraging innovative tech-
nologies, our system aims to create a dynamic and interactive educational environment conducive
to academic success.

-I-
 Acknowledgements

We would like to thank the Department of Computer Science and Engineering, Leading
University, Sylhet 3112, Bangladesh, for supporting this course work.

-II-
 Dedication

This project is dedicated to the education community, who work tirelessly to provide students
with the tools and knowledge they need to succeed in life. Teachers, professors, administrators,
and support staff all play a critical role in shaping the next generation of leaders, thinkers, and
innovators. This project is a testament to the importance of education and the value of using
technology to enhance the learning experience for students and teachers alike. By automating
certain aspects of classroom management and providing more personalized feedback, this project
seeks to alleviate some of the burdens facing educators and empower students to reach their full
potential. I am grateful for the hard work, dedication, and passion of the education community,
and I dedicate this project to their continued success in shaping the future.

-III-
 Contents

Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I
Acknowledgement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . II
Dedication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . III
Table of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . IV
List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VI

1 Introduction 1
1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Project Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

2 Background Study 4
2.1 Review Previous Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

3 Methodology 5
3.1 Model Details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3.1.1 Gemini Pro model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3.1.2 Gpt-3.5 model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3.1.3 GoogleGenerative AIEmbeddings . . . . . . . . . . . . . . . . . . . . . 8
3.2 Baseline Transformer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

4 Structural Design 10
4.1 Use case Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
4.2 Data Flow Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

-IV-
 4.3 Ai Chat system Workflow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
4.3.1 Chat with PDF . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
4.3.2 Chat with Video . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
4.3.3 MCQ generator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
4.3.4 Chat with Gemini Pro . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
4.4 Automated Attendance System . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
4.5 Find Script Copier System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

5 Result and Analysis 19

5.1 Parameters Analysis of LLMs in Recent Years . . . . . . . . . . . . . . . . . . . 19
5.2 Gemini Chat-bot Comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

6 Limitations and Future Work 22

6.1 Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
6.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

7 Conclusion 25
7.1 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

-V-
 List of Figures

3.1 Gemini model architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

3.2 Gpt model architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
3.3 GoogleGenerative AIEmbeddings architecture . . . . . . . . . . . . . . . . . . . 8
3.4 Base Transformer Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

4.1 Use case Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

4.2 Data Flow Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
4.3 chat system work flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
4.4 Chat with Gemini Pro work flow . . . . . . . . . . . . . . . . . . . . . . . . . . 16
4.5 Attendance System work flow . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
4.6 Find Copier System work flow . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

5.1 Parameters of LLMs in Recent Years . . . . . . . . . . . . . . . . . . . . . . . . 20

5.2 Parameters of LLMs in Recent Years . . . . . . . . . . . . . . . . . . . . . . . . 21

-VI-
 Chapter 1

Introduction

1.1 Background

Traditional classroom management systems often rely on manual processes, leading to in

efficiencies in communication, resource dissemination, and assignment management. With the
proliferation of digital technologies, thereâs a growing need for integrated platforms that streamline
these tasks and enhance the learning experience for both educators and students. Existing digital
solutions often lack comprehensive integration, requiring users to navigate multiple platforms for
different functions. Additionally, the shift towards remote and hybrid learning models necessitates
adaptable tools that facilitate collaboration and engagement across diverse learning environments.
The proposed system addresses these challenges by providing a cohesive platform that combines
PDF and YouTube chat-bots with assignment management functionalities.[1] By leveraging natural
language processing and multimedia integration, the system enhances accessibility to educational
resources and promotes interactive learning experiences. Moreover, the centralized assignment
management module simplifies workflow processes, enabling instructors to provide timely feed
back and students to track their progress effectively. Through innovation and integration, the
project seeks to modernize classroom management practices and foster a dynamic educational
ecosystem aligned with contemporary learning needs.

-1-
1.2 Motivation

The motivation behind the Eduverse project is to improve the classroom experience for both
students and teachers, and to make education more accessible, efficient, and effective. By leveraging
the power of machine learning, Eduverse has the potential to transform the way students learn and
interact with their teachers, ultimately leading to better educational outcomes for everyone [1]

1.3 Project Objectives

• Automated Attendance System: Develop an automated system for recording student atten-
dance using technologies such as facial recognition or QR code scanning. This system will
eliminate the need for manual attendance taking and provide accurate records for adminis-
trative purposes.

• Automated Quiz Generation: Implement an automatic quiz generation feature that gener-
ates questions based on predefined criteria such as topic, difficulty level, and question type.
This feature will save instructors time and effort in creating quizzes and ensure a variety of
questions for assessment purposes.

• Chatbot Integration: Integrate a chatbot into the platform to assist users with queries,
provide information, and guide them through the system’s features. The chatbot will use
natural language processing algorithms to understand user inputs and respond accordingly,
enhancing user experience and engagement.

• Chat with YouTube: Enable users to engage in real-time discussions and comments while
watching YouTube videos within the platform. This feature will encourage collaboration,
facilitate learning, and provide a platform for users to share insights and perspectives on the
video content.

• Chat with PDF: Implement a feature that allows users to annotate, highlight, and discuss PDF
documents within the platform. This feature will enable users to collaborate on documents,
share feedback, and exchange ideas seamlessly.

• Assignment Posting: Allow instructors to post assignments, projects, and tasks within the

-2-
 platform, specifying deadlines, instructions, and submission guidelines. Students can access
these assignments, submit their work, and receive feedback from instructors through the
platform.

• Comments on Posts: Enable users to comment on posted assignments, projects, announce-

ments, and discussions within the platform. This feature promotes collaboration, fosters
communication, and allows users to provide feedback and share insights on various topics.

• User Authentication (Login/Sign up): Implement a secure authentication system that

requires users to log in or sign up to access the platform’s features. User authentication will
protect sensitive information, ensure data privacy, and prevent unauthorized access to user
accounts.

• Automated Student Enrollment: Develop automated processes for adding students to

classrooms and courses based on enrollment lists provided by instructors or administra-
tors. This feature will streamline administrative tasks, ensure accurate student records, and
facilitate seamless enrollment procedures.

• Real-time Updates: Provide real-time notifications, alerts, and updates to users regarding
assignments, deadlines, announcements, and other relevant information within the platform.
This feature will keep users informed, engaged, and up-to-date with the latest developments
and activities.

-3-
 Chapter 2

Background Study

2.1 Review Previous Work

In the field of education, there have been several previous works that have aimed to automate
various aspects of classroom management. However, many of these solutions have been limited
in their scope and functionality, and often require significant manual input from teachers. One
previous work that is relevant to the Eduverse project is online learning platforms such as Coursera
by Oliveira at el .[2], edX, and Udemy. These platforms provide a wide range of online courses
that are accessible to students around the world, often for free or at a low cost. However, these
platforms do not provide the level of personalization and automation that Eduverse aims to achieve.
Another previous work is learning management systems (LMS) such as Blackboard and Canvas,
which provide tools for teachers to manage and deliver course materials, as well as to communicate
with students. However, these systems get from the research paper of Wang et al [3] often
require significant manual input from teachers, and do not provide the level of automation that
Eduverse aims to achieve. In terms of automated grading by Zhang et al[4], previous works include
systems such as Turnitin and Gradescope, which use machine learning algorithms to grade student
assignments and provide feedback on the paper of Kumar et al[5]. However, these systems are often
limited in their scope and do not provide the level of personalization and feedback that Eduverse
aims to achieve.

-4-
 Chapter 3

Methodology

Eduverse makes school life easier by automatically tracking attendance, creating quizzes,
chatting with YouTube and PDFs, posting assignments, and enabling comments. With its chatbot,
students can ask questions and get quick answers. Logging in and signing up is simple, allowing
students and teachers access to all updates. The system also automatically adds students to
classrooms. Everything is designed to help students and teachers stay organized and engaged. Just
think of Eduverse as your virtual school assistant, making learning fun and accessible anytime,
anywhere.

3.1 Model Details

In Eduverse, we use Gemini Pro, gpt-3.5 and GoogleGenerative AIEmbeddings to make

learning better. These tools help us understand opinions and generate ideas effectively. It’s like
having a smart assistant that knows what we need for university projects and discussions. Gemini
Pro helps us analyze opinions, while GoogleGenerative AIEmbeddings helps organize information
in a smart way.

3.1.1 Gemini Pro model

The Gemini Pro model is a powerful AI tool from Google, designed to balance capability and
efficiency. It excels at handling various tasks, making it suitable for a wide range of applications.
Compared to the larger and more complex Gemini Ultra, the Pro version is more efficient and easier

-5-
to scale for diverse uses. One of its key features is its ability to understand and process information
from both text and images, making it versatile for various needs. It’s important to note that this is
just a brief overview, and the Pro model offers a comprehensive set of functionalities beyond this
summary.

Figure 3.1: Gemini model architecture

3.1.2 Gpt-3.5 model

GPT-3.5, short for Generative Pre-trained Transformer 3.5, is a cutting-edge language model
developed by OpenAI, renowned for its ability to understand and generate human-like text. It’s like
a super-smart computer program trained on vast amounts of internet text, enabling it to understand
and generate responses to a wide range of prompts and queries.
Imagine GPT-3.5 as a giant library filled with books from all over the internet. It learns from
these books by understanding patterns and relationships between words, phrases, and concepts.
When you give it a prompt or ask it a question, it sifts through this library to find relevant information

-6-
and generate a response.
What makes GPT-3.5 special by Garcia at el.[6] is its massive size and advanced training
techniques. It’s trained on billions of parameters, which are like the building blocks of its under-
standing. These parameters allow it to capture intricate details of language and context, making its
responses more accurate and nuanced.
College students can use GPT-3.5 for various tasks like writing essays, generating ideas,
summarizing text, or even coding assistance. It’s like having a knowledgeable friend who can
help you with your homework or provide insights on complex topics. However, it’s important to
remember that while GPT-3.5 is incredibly powerful, it’s not perfect and may sometimes produce
inaccurate or nonsensical responses. Therefore, it’s essential to critically evaluate its output and
use it as a tool to augment, rather than replace, human intelligence and creativity.

Figure 3.2: Gpt model architecture

-7-
3.1.3 GoogleGenerative AIEmbeddings

GoogleGenerative AIEmbeddings is an advanced tool used for understanding and organizing

information. It employs cutting-edge artificial intelligence techniques to embed data in a mean-
ingful way, making it easier to analyze and process. Similar to how our brains connect ideas by
Miller at el.[7], this tool helps in creating associations and insights from complex information.
College students can use it for various tasks like research, data analysis, and project development.
By harnessing the power of AIEmbeddings, students can explore relationships between concepts,
streamline their workflows, and uncover valuable insights, ultimately enhancing their academic
pursuits and problem-solving skills

Figure 3.3: GoogleGenerative AIEmbeddings architecture

3.2 Baseline Transformer

The Baseline Transformer is a fundamental architecture used in natural language processing

(NLP). It’s a type of neural network designed to understand and generate human-like text. Similar
to how we interpret language, the Baseline Transformer processes words in sequences, capturing
relationships and context. It’s composed of attention mechanisms that focus on relevant parts
of text during processing. College students use Baseline Transformers for tasks like language
translation, text summarization, and sentiment analysis. By leveraging this architecture, students

-8-
can delve into complex language tasks, develop NLP applications by Clark at el.[8], and understand
the underlying principles driving modern AI technologies in language processing.

Figure 3.4: Base Transformer Model

-9-
 Chapter 4

Structural Design

The structural design chapter in our project report illustrates the system’s architecture, data
flow, and entities. It’s crucial for understanding the tour and travel agency platform’s framework
and functionality, offering insight into how components interact and data moves through the system,
guiding development and alignment with project goals.

4.1 Use case Diagram

The classroom management system serves as a digital hub where teachers and students interact
seamlessly. Teachers have the power to create virtual classrooms, just like physical ones, where
they can invite students and manage their enrollment. Within these classrooms, teachers can upload
assignments, quizzes, and other learning materials. For students, the system acts as a central point
for accessing course materials by Smith at el.[9] and assignments. They can submit their work
electronically, reducing paperwork and making the process more efficient. Moreover, students
can participate in discussions, ask questions, and collaborate with their peers within the digital
classroom environment. The system enhances communication between teachers and students.
Teachers can send announcements, reminders, and notifications to keep students informed about
upcoming deadlines or important events. Students, in turn, can seek clarification by Garcia at el.[6]
on assignments or communicate their concerns to their teachers in a structured manner. Quizzes
and assessments can be generated and administered digitally, providing instant feedback to students
and allowing teachers to track their progress more effectively.

-10-
Figure 4.1: Use case Diagram

-11-
4.2 Data Flow Diagram

The Data Flow Diagram (DFD) illustrates the flow of data within a system, exemplified here
by a classroom management system. Inputs such as student information, assignments, and grades
traverse the system, undergoing processes like enrollment, grading, and communication. Processed
data is subsequently stored in databases or files, facilitating the generation of reports, transcripts,
and notifications for teachers, students, or parents.

Figure 4.2: Data Flow Diagram

-12-
4.3 Ai Chat system Workflow

This project involves setting up the development environment, loading documentation content,
and creating a chat interface using LangChain, a language processing tool. With LangChain, the
chat interface can understand and respond to natural language input from users. Additionally, the
project includes building a basic React application to serve as the frontend interface. This interface
enables users to engage in chat conversations, access PDF documents, and watch videos within
the chat interface itself. Furthermore, the system can generate multiple-choice questions (MCQs)
based on the content of the chat or documents. This project aims to provide college students
with a convenient platform for interactive learning and communication. Users can chat with the
system, access educational materials in PDF format, watch instructional videos, and even practice
with MCQs generated by the system. Overall, it’s designed to enhance learning experiences and
facilitate knowledge acquisition in a user-friendly and engaging manner.

Figure 4.3: chat system work flow

4.3.1 Chat with PDF

Chatting with PDFs involves using advanced language models like Gemini Pro LLM and
GoogleGenerative AIEmbeddings to interact with PDF documents. These models enable students
to ask questions, seek clarifications, or request summaries directly within the chat interface. The

-13-
system processes the text content of PDF files, allowing students to engage in natural language
conversations about the document’s content. For instance, students can ask for definitions, expla-
nations of complex concepts, or key points from the PDF. The system utilizes its understanding
of the document’s content to provide relevant responses, aiding students in their learning process.
This feature enhances accessibility to educational materials and promotes interactive learning ex-
periences. College students can benefit from this functionality by accessing course materials more
conveniently and engaging in meaningful discussions about the content without having to leave the
chat interface. It streamlines the learning process and encourages active participation in academic
discussions.

4.3.2 Chat with Video

Chatting with YouTube videos involves integrating advanced language models like Gemini
Pro LLM and GoogleGenerative AIEmbeddings to interact with video content. Students can
engage in conversations about the videos they watch, ask questions, request explanations, or seek
further insights directly within the chat interface. The system processes the audiovisual content
of YouTube videos, allowing students to discuss specific moments, concepts, or ideas presented
in the video. For example, students can inquire about key points, seek clarification on complex
topics, or discuss their interpretations of the content. The system utilizes its understanding of the
video’s content to provide relevant responses, fostering deeper engagement and comprehension.
This feature enhances accessibility to educational videos and encourages active participation in
discussions about the content. College students can benefit from this functionality by accessing
instructional videos more conveniently and engaging in meaningful conversations about the topics
covered. It promotes interactive learning experiences and facilitates a deeper understanding of the
video content within the context of a chat interface.

4.3.3 MCQ generator

An MCQ generator is a tool designed to create multiple-choice questions (MCQs) automatically

or semi-automatically. It utilizes various techniques such as natural language processing (NLP)
algorithms, keyword extraction, or content analysis to generate questions based on a given text,
topic, or dataset.

-14-
 Using an MCQ generator, educators can quickly create sets of questions for quizzes, tests, or
assessments. The generator can extract key concepts, facts, or ideas from the content provided and
formulate relevant questions with multiple-choice answers.
The process typically involves: Providing the system with the relevant content, which could be
in the form of text documents, passages, or datasets. The system analyzes the content, identifies
important information, and determines suitable topics for generating questions. Based on the
analysis, the generator formulates multiple-choice questions with distractors (incorrect options)
and correct answers. Users may have options to customize the difficulty level, number of questions,
or specific topics they want to focus on. The generator produces a set of MCQs ready for use in
assessments or learning activities.
MCQ generators save time for educators and offer a standardized approach to creating assess-
ment materials. They also ensure consistency in question formulation and can be valuable tools in
educational settings, particularly in online learning environments where frequent assessments are
conducted.

4.3.4 Chat with Gemini Pro

Chatting with Gemini Pro presents an exciting opportunity for college students to converse with
a cutting-edge AI model developed by Google AI. By sending text prompts, students can engage in
dynamic conversations, where Gemini Pro utilizes its extensive knowledge of language by Chen at
el.[10] to generate responses. The model adapts its answers based on previous messages, allowing
for fluid and contextually relevant interactions. With features like viewing chat history and saving
conversations, students can explore the depths of dialogue and creativity. Gemini Pro enables
students to ask questions, discuss diverse topics, and experiment with different writing styles,
fostering curiosity and intellectual exploration. Accessible through platforms like Google Cloud
Vertex AI Studio and community-developed Python applications by Lee at el.[11], this innovative
tool offers insights into the workings of large language models. However, students should bear
in mind that while Gemini Pro allows interaction, its responses may occasionally lack absolute
accuracy or completeness, given its ongoing developmental stage.

-15-
 Figure 4.4: Chat with Gemini Pro work flow

4.4 Automated Attendance System

The Automated Attendance System operates seamlessly through the integration of Python,
OpenCV, and MongoDB, each contributing essential functionalities:
Python serves as the primary programming language, offering versatility and ease of integration
with various libraries and frameworks. It orchestrates the system’s functionalities, coordinating
data flow between components and facilitating interaction with external systems.
OpenCV, an open-source computer vision library, plays a pivotal role in face detection and
recognition. Leveraging sophisticated algorithms, OpenCV analyzes by Wang at el.[12] video
streams or images captured by cameras to identify faces accurately. Through image processing
techniques, it detects facial features, extracts pertinent information, and matches them against
existing records within the system.
MongoDB acts as the backend database, storing attendance records, user profiles, and system
configurations. Its flexible NoSQL architecture accommodates unstructured data, making it ideal
for handling diverse data types generated by the attendance by Jones at el.[13] system. MongoDB’s
scalability and robust querying capabilities enable efficient data retrieval and management, ensuring

-16-
seamless integration with other components.
The system’s workflow begins with the initialization and configuration of hardware components
and software dependencies. As individuals enter the designated area, cameras capture their images,
which are processed in real-time by OpenCV for face detection and recognition. Upon successful
identification, the system interacts with MongoDB to record attendance entries, associating them
with corresponding timestamps and user profiles.
To enhance system reliability and performance, error handling mechanisms mitigate challenges
such as lighting variations, occlusions, and facial by Gupta at el.[14] obstructions. Continuous
optimization and monitoring ensure consistent operation, minimizing false positives and negatives
in attendance tracking.
Administrators interact with the system through a user-friendly interface, accessing attendance
reports, monitoring real-time data, and managing system configurations. Customizable features
empower administrators to tailor the system to their specific requirements, fostering adaptability
and usability.

Figure 4.5: Attendance System work flow

-17-
4.5 Find Script Copier System

The Student Copier System is an innovative platform designed to streamline academic integrity
processes while ensuring fairness and originality in student submissions. At its core, the system
integrates a student script utilizing FastAPI, providing a seamless and efficient interface for users.
Leveraging the powerful functionalities of the difflib PyPI library, the system employs cutting-edge
algorithms to analyze text and identify similarities between documents.
Powered by MongoDB, the system utilizes a sophisticated library for efficiently locating in-
stances of copied content within a group of submissions. Employing pairs similarity techniques, it
can detect similarities between documents with high accuracy and reliability.
Furthermore, the system seamlessly integrates with Google’s Apply URL, allowing for ef-
ficient scraping of relevant content using BeautifulSoup. With stability as a top priority, the
system employs robust mechanisms to handle diverse data formats, ensuring compatibility with a
wide range of input sources. Utilizing cosine similarity algorithms, it compares documents in a
symbolic format, enabling precise identification of copied passages. Additionally, the system effec-
tively navigates through HTML tags, ensuring thorough examination of content while maintaining
performance and accuracy.

Figure 4.6: Find Copier System work flow

-18-
 Chapter 5

Result and Analysis

In my project, result analysis entails evaluating the performance and effectiveness of various
components, including the automated attendance system and chat interfaces with Gemini Pro and
other models. This involves assessing the accuracy of attendance records, the quality of interactions
with language models, and user engagement with additional features like chat with PDFs, YouTube
videos, and MCQ generation. The analysis includes comparing recorded attendance against ground
truth data, evaluating response accuracy and user satisfaction in chat interactions, and examining
user engagement metrics for additional features. Insights from the analysis inform improvements,
identify areas of strength and weakness, and guide further development and refinement of the
project. By conducting thorough result analysis, your project can ensure functionality optimization
and enhance user experience, ultimately meeting the intended objectives effectively.

5.1 Parameters Analysis of LLMs in Recent Years

Natural Language Processing (NLP) is a vital area of artificial intelligence focusing on com-
puters’ interaction with human languages. Language Modeling (LM) is a fundamental NLP task,
involving statistical models to analyze and generate natural language. Recent advancements in
LM have seen the rise of Large Language Models (LLMs), like the 175B-parameter GPT-3 and
the 541B-parameter PaLM model. These models, trained on vast datasets using deep learning
techniques, excel in various NLP tasks like machine translation, sentiment analysis, and dialogue
systems.

-19-
 LLMs demonstrate remarkable abilities such as few-shot and zero-shot learning, adapting
quickly to new tasks with minimal examples and generating reasonable responses to unseen
prompts. ChatGPT, powered by GPT LLMs, showcases exceptional dialogue capabilities and
finds applications in customer service, language tutoring, and personal assistance.
LLMs hold promise for revolutionizing fields like education, healthcare, and business due to
their profound impact on natural language processing. Their scalability and performance across
diverse NLP tasks highlight their potential to enhance human-computer interactions, automate
processes, and unlock new possibilities for innovation and productivity in various domains.

Figure 5.1: Parameters of LLMs in Recent Years

5.2 Gemini Chat-bot Comparison

Google’s unveiling of Gemini, its new AI chat-bot, caused a significant 5 percentage surge in
its stock price. Gemini is touted as an upgrade to Bard, Google’s previous AI model. The aim
is clear: to rival OpenAI’s GPT-4, which remains one of the top contenders in the AI chat-bot
arena. However, Google has yet to release comparisons between Gemini and GPT-4, leaving

-20-
users curious about its capabilities. According to Wired.com, Gemini Pro, the base version,
is available for free, contrasting with GPT-4, which requires a subscription for access. While
Gemini Ultra, an even more advanced iteration, is anticipated for 2024, details about its release
and subscription plans remain scarce. Wired.com offers insights into using Gemini effectively,
emphasizing its potential pitfalls and reminding users to remain cautious about chat-bot accuracy.
Gemini’s current limitation lies in its English-only support, posing a challenge for non-English
speakers. However, its emergence underscores Google’s prowess in AI development, potentially
influencing its stock performance positively. Despite the hype, users must remain patient as they
await more information and real-world comparisons to determine Gemini’s true capabilities.

Figure 5.2: Parameters of LLMs in Recent Years

-21-
 Chapter 6

Limitations and Future Work

6.1 Limitations

1. Technical Constraints:

The effectiveness of the chatbots, particularly the YouTube chatbot and PDF chatbot, may
be limited by technical constraints such as internet connectivity issues, compatibility with
different devices, and platform stability.

2. User Adoption Challenges:

While the introduction of new technologies like chatbots and LLM models may enhance the
learning experience, there may be resistance to adoption among teachers and students who
are accustomed to traditional methods.

3. Privacy and Security Concerns:

The collection and storage of student data, particularly in the context of generating MCQ
questions using LLM models, raise privacy and security concerns.

4. Bias and Accuracy of AI Models

The accuracy and fairness of the LLM models used for generating MCQ questions may be
influenced by biases in the training data or limitations of the underlying algorithms. Users
may encounter instances where the generated questions are biased or inaccurate, leading to
inconsistencies in assessment and potential disparities in student performance evaluation.

-22-
 5. Scalability and Maintenance:

As the user base grows and the platform’s usage increases, scalability and maintenance
become critical challenges. Scaling up the infrastructure to accommodate a larger user load,
ensuring high availability.

6. Lack of Customization:

The platform may lack sufficient flexibility and customization options to accommodate
diverse educational contexts and pedagogical approaches. Teachers and institutions may
require features tailored to their specific needs, such as customizable grading criteria, as-
signment templates, or integration with external learning resources. Striking a balance
between standardization and customization will be essential to meet varied user require-
ments effectively.

6.2 Future Work

Analysis Dashboard Tool:

In the future, analysis dashboard tools will likely become more intuitive and comprehensive,
offering educators and administrators deeper insights into various aspects of student performance,
engagement, and learning outcomes. These tools may leverage advanced data visualization tech-
niques, machine learning algorithms, and predictive analytics to identify patterns, trends, and areas
for improvement in real-time. Additionally, they may integrate seamlessly with other educational
technologies to provide a holistic view of the learning environment.

Bias and Accuracy of AI Models:

As AI continues to play a significant role in education, addressing bias and improving the
accuracy of AI models will become paramount. Future advancements may involve the development
of more sophisticated algorithms that are capable of identifying and mitigating bias in data sources,
training processes, and decision-making algorithms. Additionally, there will likely be increased
emphasis on transparency, accountability, and ethical considerations in the design and deployment
of AI-powered educational systems.

-23-
Privacy and Security Concerns:

With the proliferation of data-driven educational technologies, privacy and security concerns
will remain a top priority in the future. Educational institutions and technology providers will need
to implement robust data protection measures, encryption protocols, and access controls to safe-
guard sensitive student information. Furthermore, there may be greater emphasis on empowering
students and educators with tools and resources to control their personal data and understand the
implications of data collection and usage practices.

AI Generated Text Copier:

In the future, AI-generated text copiers may become more sophisticated, raising concerns
about academic integrity and plagiarism detection. Educational institutions will need to invest
in advanced plagiarism detection tools that are capable of identifying content generated by AI
algorithms. Additionally, there may be efforts to promote digital literacy and ethical writing
practices among students to prevent misuse of AI technology for academic dishonesty.

Real-time Communicating System:

The future of real-time communication systems in education will likely be characterized by

seamless integration across multiple platforms and devices. These systems may leverage AI-
powered chatbots and natural language processing algorithms to facilitate instant communication
between students, educators, and administrators. Furthermore, there may be greater emphasis
on accessibility and inclusivity, ensuring that real-time communication systems are designed to
accommodate diverse learning needs and preferences.

Automated Marking System:

Automated marking systems will continue to evolve in the future, offering educators more effi-
cient and accurate assessment tools. These systems may incorporate machine learning algorithms
to analyze student responses, provide personalized feedback, and identify areas for improvement.
Additionally, there may be increased integration with learning management systems and digital
assessment platforms, enabling seamless tracking of student progress and performance over time.

-24-
 Chapter 7

Conclusion

7.1 Discussion

In conclusion, the proposed classroom management system represents a significant step forward
in enhancing educational experiences through innovative technology integration. By leveraging
PDF and YouTube chatbots, streamlined assignment management, and AI-driven MCQ generation,
the platform addresses key challenges in communication, engagement, and assessment within
traditional classroom settings. Despite its promising features, the system’s implementation may
face technical, user adoption, and privacy challenges. However, with careful consideration of
these limitations and proactive measures to address them, the platform holds immense potential
to transform teaching and learning paradigms. Moving forward, it is imperative to prioritize user
feedback, refine AI models, and ensure compliance with privacy regulations to foster a supportive
and inclusive learning environment. By embracing continuous improvement and collaborative
efforts, the platform can evolve into a versatile tool that empowers educators, engages students, and
cultivates lifelong learning skills. Ultimately, the success of the classroom management system
hinges on its ability to adapt to evolving educational needs and inspire positive learning outcomes
for all stakeholders.

-25-
 References

[1] Y. Li and H. Oliveira, “Research on classroom practice,” Available in:

https://link.springer.com/chapter/10.1007/978-3-319-12688-34 6, 2015.

[2] . C. N. Wang, C., “Enhancing student engagement through chatbot integration in online
learning environments: An exploratory study.” Computers Education, 136, 1-11., 2019.

[3] . H. S. Kizilcec, R., “Attrition and achievement gaps in online learning.” Proceedings of the
Second (2015) ACM Conference on Learning @ Scale, 57-66., 2015.

[4] . Z. J. Zhang, J., “A survey on chatbots in education: From intelligent tutoring systems to
conversational agents.” Computers Education, 157, 1-15., 2020.

[5] . S. M. Kumar, A., “Role of artificial intelligence in education: A review study.” Journal of
Critical Reviews, 7(18), 104-111., 2020.

[6] . R. P. Garcia, M., “A systematic review of chatbots in education: Trends and opportunities,”
International Journal of Educational Technology in Higher Education, 18(1), 1-25., 2021.

[7] . J. L. Miller, K., “The use of chatbots in teaching and learning: A review of the literature,”
Journal of Computing in Higher Education, 33(1), 21-40., 2021.

[8] . T. L. Clark, R., “Chatbots in higher education: A systematic literature review,” British
Journal of Educational Technology, 52(5), 2424-2447., 2021.

[9] . B. A. Smith, J., “Chatbots in education: A literature review,” Journal of Interactive Learning
Research, 31(2), 165-188., 2019.

-26-
[10] . H. Y. Chen, Y., “Enhancing personalization in education with chatbots: A systematic
review,” Journal of Educational Computing Research, 58(3), 501-529., 2021.

[11] . K. M. Lee, J., “The impact of chatbots on student engagement and learning outcomes:
A meta-analysis,” Educational Technology Research and Development, 69(4), 2021-2042.,
2019.

[12] . L. S. Wang, H., “Adapting chatbots for educational purposes: A review of existing literature,”
British Journal of Educational Technology, 52(5), 2307-2323., 2021.

[13] . S. T. Jones, S., “Exploring the role of chatbots in higher education: A case study approach,”
Journal of Educational Technology Systems, 48(2), 191-211., 2020.

[14] . S. A. Gupta, R., “The future of chatbots in education: Opportunities and challenges,”
International Journal of Emerging Technologies in Learning, 16(13), 84-100., 2021.

-27-