GENERATIVE AI CHATBOT

A major project report submitted in partial fulfillment of

the requirement for the awarded degree of
Bachelor of Technology in

Computer Science and Engineering

Submitted by

Anirudh Farwaha (201446)

Jyotirmay Verma (201157)

Under the guidance and supervision of

Mr. Faisal Firdous

Mr. Ramesh Narwal

Department of Computer Science & Engineering

and Information Technology

Jaypee University of Information Technology,

Waknaghat 173234, Himachal Pradesh, INDIA
 DECLARATION
We hereby declare that the work presented in this report entitled ‘Generative AI
Chatbot’ in partial fulfillment of the requirements for the award of the degree of
Bachelor of Technology in Computer Science & Engineering submitted in the
Department of Computer Science & Engineering and Information Technology,
Jaypee University of Information Technology, Waknaghat is an authentic record
of our own work carried out over a period from August 2023 to May 2024 under
the supervision of Mr. Faisal Firdous (Assistant Professor (SG), Department of
Computer Science & Engineering and Information Technology).

The matter embodied in the report has not been submitted for the award of any
other degree or diploma.

Mr. Anirudh Farwaha Mr. Jyotirmay Verma

(201446) (201157)

This is to certify that the above statement made by the candidate is true to the
best of my knowledge.

Mr. Faisal Firdous

Assistant Professor
Department of Computer Science & Engineering and Information
Technology Jaypee University of Information Technology

Mr. Ramesh Narwal

Assistant Professor
Department of Computer Science & Engineering and Information
Technology Jaypee University of Information Technology

I
 CERTIFICATE
This is to certify that the work which is being presented in the project report
titled ‘Generative AI Chatbot’ in partial fulfillment of the requirements for the
award of the degree of B.Tech in Computer Science And Engineering and
submitted to the Department of Computer Science And Engineering, Jaypee
University of Information Technology, Waknaghat is an authentic record of work
carried out by “Anirudh Farwaha (201446)”, “Jyotirmay Verma (201157)”
during the period from August 2023 to May 2024 under the supervision of Mr.
Faisal Firdous, Assistant Professor (SG), Department of Computer Science and
Engineering, Jaypee University of Information Technology, Waknaghat.

Submitted by:

Mr. Anirudh Farwaha Mr. Jyotirmay Verma

(201446) (201157)

The above statement made is correct to the best of my knowledge

Mr. Faisal Firdous

Assistant Professor
Department of Computer Science & Engineering and Information
Technology Jaypee University of Information Technology

Mr. Ramesh Narwal

Assistant Professor
Department of Computer Science & Engineering and Information
Technology Jaypee University of Information Technology

II
 ACKNOWLEDGEMENT

Firstly, we express our heartiest thanks and gratefulness to almighty God for his
divine blessings that makes it possible for us to complete the project work
successfully.

We are really grateful and wish our profound indebtedness to our Supervisor Mr.
Faisal Firdous(Assistant Professor), Department of CSE Jaypee University of
Information Technology,Wakhnaghat). Deep Knowledge & keen interest of our
supervisor in the field of our project “Generative AI Chatbot” helped us to
carry out this project. Her endless patience, scholarly guidance, continual
encouragement, constant and energetic supervision, constructive criticism,
valuable advice, reading many inferior drafts and correcting them at all stages
has made it possible for us to complete this project.

We would like to express our heartiest gratitude to Mr. Faisal Firdous(Assistant

Professor, Department of CSE), for her kind help to finish our project.

We would also generously welcome each one of those individuals who have
helped us straightforwardly or in a roundabout way in making this project a win.
In this unique situation, We might want to thank the various staff individuals,
both educating and non-instructing, which have developed their convenient help
and facilitated our undertaking.

Finally, We must acknowledge with due respect the constant support and
patience of our parents.

Mr. Anirudh Farwaha Mr. Jyotirmay Verma

(201446) (201157)

III
 ABSTRACT

Generative synthetic intelligence (AI) chatbots have revolutionized

conversational interfaces by offering state-of-the-art capabilities, including
post-processing and natural language understanding. It highlights key
advancements in generative AI chatbots and examines their impact on
transforming human-computer interaction. Generative AI chatbots leverage
advanced natural language processing (NLP) techniques, notably deep learning
models like GPTs (Generative Pre-Learn Transformer 3), enabling them to
comprehend and respond to contextual cues. Unlike rule-based or retrieval-based
chatbots, generative models have the ability to generate new coherent responses
and accommodate new dynamic rhythms of conversation.

Ethical considerations such as bias, disinformation, and responsible use are also
discussed in relation to generative AI chatbots. It reviews current research and
development initiatives aimed at improving the accountability and transparency
of these models. Furthermore, it explores the positive applications of generative
AI chatbots in various sectors, including customer service, healthcare, education,
and entertainment. Empirical studies and case examples illustrate how these
chatbots can simplify communication, enhance user satisfaction, and boost
productivity.

Looking forward, It anticipates advancements in generative AI chatbots,

including enhancements in multi-modal capabilities, emotional intelligence, and
ongoing efforts to address ethical challenges. As these intelligent conversational
agents continue to evolve, they are poised to shape a new era of seamless and
natural communication, impacting human-device interactions.

IV
 Table Of Contents

Declaration I
Certificate II
Acknowledgement III
Abstract IV

Chapter 01 1
1.1 Introduction 1
1.2 Problem Statement 2
1.3 Objectives 3
1.4 Significance and Motivation 4
1.5 Organization of Project Report 5
Chapter 02 6
2.1 Overview of Relevant Literature 6
2.2 Key Gaps in the Literature Survey 11
Chapter 03 12
3.1 Requirements and Analysis 12
3.2 Project Design and Architecture 15
3.3 Data Preparation 17
3.4 Implementation 22
3.5 Key Challenges 40
Chapter 04 34
4.1 Testing Strategy 34
1.2 Test Cases and Outcomes 36
Chapter 05 42
5.1 Results 42
5.2 Comparison with Existing Solutions 45
Chapter 06 48
6.1 Conclusion 48
6.2 Future Scope 49
References 50

V
 List Of Figures

Fig No. Figure Name Pg No.

1.1 Chatbot Flowchart 15

1.2 Dialogflow working 16

1.3 User Authentication Flowchart 16
1.4 NLP w.r.t. NLP Diagram 20
1.5 Selected competition 20
1.6 - 1.12 Implementation 22-26
1.13 - 1.15 Backend Database 26-27
2.1 - 2.5 Code Snippets 28-30
3.1 - 3.6 Testing 37-40
3.7 ngrok localhost to server 40
3.8 Localhost Expess.js Server 41
4.1 Home Page UI 42
4.2 website with chatbot activated 43
4.3 Payment gateway 43
4.4 Payment Successful 44
4.5 Menu and cart total 44

VI
 List Of Tables

S No. Figure Name Page

1. Literature survey 7-10

VII
 Chapter-01 INTRODUCTION
1.1 Introduction

Conversational AI is being revolutionized by means of generative artificial

intelligence (AI) chatbots, which might be converting the manner humans
communicate with shrewd structures. Powered through superior deep mastering
fashions inclusive of GPT-3, those chatbots pass beyond conventional rule-based
or retrieval-based techniques with the aid of demonstrating an outstanding
capacity to recognize and produce contextually relevant responses.The capability
of generative AI chatbots to be responsive towards past the pre- programmed
and instead grasp the power of large language models which have been
previously skilled on a number of datasets is what makes them so effective. They
are able to realize the originality of human language as an end result, capturing
context and nuances for more organic and lively encounters.

The architecture and education techniques that assist generative AI chatbots are
examined on this research record. The flexibility and customization potential of
those chatbots emerged as obvious in the course of their pre-schooling segment,
which worried large datasets, and in the course of the best-tuning section, which
involved mission-specific edition. This study specializes in the hard balancing
act of capturing widespread linguistic patterns and customizing responses to
particular contexts. This discovery also has an essential impact for the ethical
use of generative AI chatbots. Strong ethical frameworks are important for the
creation and application of these technologies, as can be impactful in the wrong
way which can be biased and can also be because of disinformation, and
responsible deployment. We look at real-global applications in a whole lot of
sectors, along with healthcare, education, enjoyment, and customer support, to
illustrate the concrete effects of generative AI chatbots. We exhibit how those
chatbots enhance consumer studies, expedite communication, and boom
operational performance with case studies and actual-world scenarios.

As we start our research, our goals are to understand the country of generative
AI chatbots now and to predict wherein they will move inside the destiny. As we
make our manner through this revolutionary terrain of smart conversational bots,
potential trends in multi- modal capabilities, emotional intelligence, and
persevering with efforts to deal with ethical issues might be vital focal points. In
the pages that comply with, we explore the generative AI chatbots' ramifications
and minute capabilities, revealing their importance in influencing the following
phase of easy and realistic human-gadget interactions.

1
1.2 Problem Statement

In a rapidly evolving landscape of online food ordering our goal are to map out
the modern generative AI chatbots and forecast their destiny trajectory with
suggestions and easy to convey and place order. Prospective likelihood in
multi-modal talents, emotional intelligence and tenacity in addressing ethical
dilemmas can also prove to be vital focuses as we navigate this with the help of
sensible conversational bots. In the following pages we search into the
consequences and subtleties of generative AI chatbots, highlighting their
importance in shaping the following stage of simple and realistic human-tool
interactions.
The existing problem lies inside the lack of ability of conventional chatbots to
apprehend and reply dynamically to personal queries, possibilities for food and
the evolving context of a communique. They have a tendency to form frustration
in the direction of the era because of lack of interactions and conflict with an
inflexible machine who fails to conform their particular preferences or handle
complex requests efficiently.
It becomes critical to create a generative AI chatbot designed especially for a
meal ordering internet site so as to address this issue. The objective is to create a
conversational agent which can comprehend spoken language, cope with
complicated consumer requests, and reply in a manner that makes feel for the
given context. This chatbot seeks to enhance the user revel in with the aid of the
use of present-day device gaining knowledge of and natural language processing
strategies to handle order changes readily, make personalized food tips, and
create a more thrilling and person-pleasant online ordering platform. In the
specific placing of an internet restaurant ordering platform, this looks to shut the
space among patron expectations and the constraints of present chatbot
technology.

2
1.3 Objectives

The primary objective of this project is to develop an generative AI

chatbot to help users with their queries ordering food and saving
business money and workforce.

● Developing a generative AI chatbot to increase user engagement by

providing dynamic and contextually relevant interactions during the
food ordering process.
● Implementing a recommendation system within the chatbot to analyze
user preferences and offer personalized food suggestions, making a
more enjoyable experience for users.
● Making it capable to understand and alter to the converting context for
user interactions, guaranteeing a smooth and natural conversation to
raise consumer happiness.
● Including the capabilities that permit the chatbot to effectively and
person pleasant to cope up with order revisions, which includes
additions, substitutes, or cancellations.
● Simplifying the system of choosing food with the aid of enabling
customers to talk with the chatbot using herbal language processing.
● Build the chatbot with web and mobile accessibility in mind to make it
available on a whole lot of systems and provide a consistent,
consumer- friendly experience for a huge range of users.

3
1.4 Motivation And Significance

The motivation and significance for building AI chatbots are multifaceted,

encompassing technological advancement, enhanced user experiences, and
streamlined business operations. At the core of this drive is the aspiration to
create more efficient and accessible communication channels that mimic
human-like interaction. One primary motivation is the continual advancement of
AI technologies, particularly in natural language processing (NLP) and machine
learning. These advancements enable chatbots to understand and respond to
human language more effectively, making them increasingly valuable for a wide
range of applications. For businesses, AI chatbots offer the opportunity to
automate customer service, sales, and support functions, reducing costs and
improving efficiency.

Another significant factor driving the development of AI chatbots is the desire to

enhance user experiences. By providing users with personalized, timely, and
relevant interactions, chatbots can improve customer satisfaction and
engagement. For example, in the healthcare sector, chatbots can provide patients
with information about their conditions or help them schedule appointments,
improving overall healthcare experiences. Furthermore, AI chatbots are seen as a
way to streamline business operations and increase productivity. By automating
repetitive tasks and providing quick access to information, chatbots can help
employees work more efficiently. For example, in a corporate setting, chatbots
can assist with onboarding processes, answer common HR-related questions, or
help employees find relevant information within a company's internal systems.

In addition to these practical benefits, AI chatbots also hold significance in

advancing the field of artificial intelligence itself. They serve as a platform for
testing and refining new AI technologies, driving further innovation in the field.
Moreover, the ethical considerations surrounding AI chatbots, such as ensuring
fairness and transparency in their decision-making processes, contribute to the
broader discourse on AI ethics and governance. In conclusion, the motivation
and significance for building AI chatbots stem from the desire to leverage
technological advancements to improve user experiences, streamline business
operations, and advance the field of artificial intelligence. As these technologies
continue to evolve, AI chatbots are poised to play an increasingly important role
in shaping the future of human-computer interaction.

4
1.5 Organization of Project Report

This project report is made in a way to provide a complete understanding of the

Generative AI chatbot, discussing its objectives, methodology, implementation,
challenges, results, and future scope. The report is organized into six chapters,
each dealing with a specific aspect of the project:

Chapter 1: Introduction
This chapter deals with introduction to the Generative AI chatbot, providing a
brief overview of its purpose, significance, and motivation. It highlights the
objectives and problem statement of the project

Chapter 2: Literature Survey

This chapter presents an overview of relevant literature and key gaps in the
literature studied during the making of Generative AI chatbot, covering existing
research and developments in the field of chatbots and technologies.

Chapter 3: System Development

This chapter delves into the requirements and analysis phase of the project. It
gives a detailed description of Project Design and Architecture, describes the
Data preparation process and Implementation of the project along with the
screenshots of the code snippets and finally Key challenges are addressed that
were faced during the project.

Chapter 4: Testing
This chapter describes the testing methodology used in testing of the chatbot,
explaining the test cases and outcomes for various components of the code of
the generative AI chatbot.

Chapter 5: Results and Evaluation

This chapter discusses the key findings of the project and their interpretation
along with a snapshot of the results and evaluation metrics for intent detection,
webhooks, fulfillment and GPT integration and comparisons are performed.

Chapter 6: Conclusion and Future Scope

This chapter concludes the project report by summarizing the key achievements,
contributions and limitations of the Generative . It also discusses potential
future scope and enhancements for the application.

5
 Chapter-02 Literature Survey
2.1 Overview of relevant literature

A literature survey on generative AI chatbots reveals a rapidly evolving field

with significant research contributions and advancements. Researchers have
explored various aspects of generative AI chatbots, including their design,
development, applications, and ethical considerations.One key area of focus in
the literature is the underlying technologies and methodologies used to develop
generative AI chatbots. Studies have examined the use of deep learning models,
such as GPTs (Generative Pre-Training Transformers), to enhance the chatbots'
natural language processing capabilities. Additionally, researchers have
investigated techniques for improving the chatbots' ability to generate coherent
and contextually relevant responses, such as fine-tuning pre-trained models and
incorporating user feedback.Another important aspect of research in this field is
the application of generative AI chatbots in various domains. Studies have
demonstrated the effectiveness of these chatbots in customer service, healthcare,
education, and entertainment. Researchers have highlighted how generative AI
chatbots can streamline communication, enhance user satisfaction, and improve
productivity in these domains.

Ethical considerations have also been a significant focus in the literature survey.
Researchers have discussed the potential ethical issues related to bias,
disinformation, and responsible use of generative AI chatbots. Efforts have been
made to address these issues through research and development initiatives aimed
at improving the accountability and transparency of these chatbots.Overall, the
literature survey highlights the growing interest and research activity in the field
of generative AI chatbots. Researchers continue to explore new technologies,
applications, and ethical considerations to further advance the field and harness
the potential of these chatbots for a wide range of applications.

6
Literature Review Table

S Paper Title Journal & Tools and Limitations

No. Conference Techniques
Year
1. "A Survey of Various chatbot Analyzed chatbot Limited focus
Chatbot frameworks, performance in on recent
Systems Loebner Prize Loebner Prize advancements
through a data competitions, and real-world
Loebner Prize highlighting applications of
Competition improvements chatbots
Lens"[5]

2. "BERT for BERT, Dialog Developed a Limited

Chatbot datasets model for natural discussion on
Development: A language chatbot
Survey"[6] inference in response
dialogue systems, generation
improving
chatbot
understanding

3. "Dialogue SNLI, Provides an Lacks a deep

Natural MultiNLI, and overview of the focus on
Language In-house evolution of NLP chatbot-
Inference" datasets models, including specific
[7] their impact on applications
chatbots

4. "A Review of the Various neural Canny edge Lacks a

Neural models, corpora detection and deep focus
History of hough lines on
Natural transform chatbot specific
Language application
Processing"
[8]

5. "Chatbots: A Various chatbot Offers a broad Limited

Survey"[9] platforms survey of chatbot technical depth
technologies, in discussing
their use cases, NLP
and challenges techniques

7
S No. Paper Title Journal & Tools and Limitations
Conference Techniques
Year
6. "GPT-3: A OpenAI GPT-3, Examines the Concerns about
Game-Changer In- capabilities of ethical use,
in Chatbot house data GPT-3 in chatbot cost, and
Development?" development, control over
[10] highlighting its GPT-3
language
generation

7. "Seq2Seq Seq2Seq, Proposes Limited

Models for Dialogue Seq2Seq models discussion on
Task-Oriented Datasets for task-oriented handling multi-
Dialog"[11] dialog systems, turn dialogues
showing
improvements in
response quality

8. "An overview of Various Proposes Focuses more

Machine emotion Seq2Seq models on emotion
learning in detection tools, for task-oriented detection than
chatbot" emotion dialog systems, on generating
[13] datasets showing empathetic
improvements in responses
response quality

9. "Designing the Multilingual Discusses the Primarily

User Interface chatbot challenges and focuses on user
for Multilingual platforms considerations interface
Chatbots"[14] for designing design and user
multilingual experience
chatbots rather than
NLP
techniques

10. "Advancements Dialog Flow Improved user Limited

in Generative AI engagement support for
for NLP highly
Chatbots" technical
[15] queries

8
S No. Paper Title Journal & Tools and Limitations
Conference Techniques
Year
11. "Deep Learning GPT-3 Achieved state- of-the-art
Approaches for of-the-art performance
NLP-based performance High
Chatbots" computational
[16] resource
requirement

12. "Context-aware Rasa Increased Challenges in

NLP Chatbots: contextual handling
A Comparative understanding contextual
Study" shifts
[17]

13. "Exploring Rich Google BERT Google BERT Limited

Messaging in standardization
NLP Chatbot in rich
Design" messaging
[19] formats

14. "Natural OpenAI GPT-4 Enhanced natural Risk of biased

Language language language
Generation in generation generation
Conversational
Agents"
[20]

15. "Multilingual Language Expanded user Translation

NLP Chatbots: Embeddings reach across inaccuracies in
Challenges and languages complex
Solutions" phrases
[21]

9
S Paper Title Journal & Tools and Limitations
No. Conference Year Techniques
16. "User-Centric User Feedback Positive Challenges in
Evaluation of Analysis feedback meeting
NLP Chatbot on diverse user
Experiences" conversational expectations
[22] style

17. "Privacy and End-to-End Enhanced user Potential

Security Encryption data protection impact on
Considerations chatbot
in NLP performance in
Chatbots" real-time
[23]

18. "Enhancing Machine Learning Machine Dependency on

User Recommender Learning user data for
Engagement Systems Recommender effective
through Systems personalization
Personalization"
[23]

19. "Efficient Distributed Reduced Complex

Training Training training infrastructure
Strategies for time for large requirements
Large-Scale models
Chatbots"
[24]

20. "Conversational UX/UI Prototyping Intuitive and Potential

UX Design for user-friendly challenges in
NLP Chatbots" interfaces adapting to
[25] diverse user
preferences

10
2.2 Key Gaps in the Literature

1. Insufficient Attention to Post-Competition Evolution: The paper may

have overlooked discussing the post-competition evolution of chatbots,
omitting crucial information on how these systems enhance themselves in
response to feedback from competitions and advancements in technology.
Additionally, there appears to be a gap in the document's analysis of the
metrics and standards used in the Loebner Prize Competition regarding their
real-world applicability. This gap could explain how effectively these
systems perform in non-competitive, everyday situations. Furthermore, with
the growing concerns about fairness and transparency in AI, it is important to
note that the study may not adequately address ethical issues and potential
biases inherent in chatbot systems.

2. Comprehensive Analysis of Chatbot Studies: This study provides a

thorough examination of chatbots, highlighting key areas where research may
be lacking. It identifies gaps in coverage of ethical issues, such as bias and
privacy, as well as limitations in analyzing chatbots' multimodal capabilities
and user experience. Additionally, the study suggests a need for more
comprehensive historical analysis of natural language processing (NLP)
pre-deep learning era and better inclusion of non-English language
processing in NLP literature. The paper emphasizes the importance of
understanding the limitations of neural NLP models and integrating
technologies like machine learning and NLP to enhance chatbot functionality
across various sectors.

3. Analysis of Chatbot Ethics, Multimodal Features, Interaction

Quality, and GPT-3 Impact: This study provides a thorough analysis of
chatbot development and technology, highlighting gaps in the coverage of
ethical issues such as bias, privacy, and AI use. It also identifies a lack of
research on chatbots' multimodal skills and insufficient reporting on user
experience. Additionally, the study explores the significant impact of
OpenAI's GPT-3 on chatbot development, delving into its architecture,
capabilities, and real-world applications.

4. Multilingual Chatbot Interface Design: This study identifies several

key gaps in the examination of multilingual chatbot interface design. Firstly,
there may be flaws in considering cultural sensitivity, which is crucial for
promoting user happiness and engagement across cultures. Secondly, there is
a need for the development of reliable assessment criteria specifically
tailored for evaluating user experience in multilingual chatbot interfaces.

11
 Chapter-03 System Development
3.1 Requirements and Analysis

Language Used: Python 3.11.2

Technical Requirements:

● A computer with at least 4GB of RAM and a multicore processor

● Internet Connection

Software:

● Python 3.5 or higher

● Visual Studio Code or any other code editor

Prerequisites:

● Dialogflow - cloud-based natural language processing service provided

by
● Google
● FastAPI - FastAPI is a modern, fast (high-performance), web framework
for
● building APIs with Python 3.7+ based on standard Python type hints.
● MySQL Workbench - To interact with the MySQL database from your
FastAPI
● application, you'll need a MySQL connector library.
● Uvicorn - Uvicorn is an ASGI server, the interface used by FastAPI.
● Ngrok - Ngrok is a tool that creates a secure tunnel to your localhost,
making
● Your FastAPI application is accessible from the internet.

12
Analysis:

● Menu Complexity Understanding: Understanding menu intricacies,

including variations and dynamic changes, guides the chatbot's
menu-handling capabilities for seamless navigation.
● Language Adaptation: To enhance linguistic flexibility, the study
should examine regional language differences, slang, and colloquialisms
related to food choices.
● Integration Challenges: Evaluation should address current website
infrastructure, potential integration issues with the chatbot, compatibility
with various platforms, scalability, and other technical constraints.
● Security and Privacy Compliance: A thorough examination of security
protocols and privacy compliance is crucial, including compliance with
data protection laws, secure user authentication, and data encryption,
given the sensitivity of user data in food delivery services.

13
Functional Requirements

● User Input Handling: Ability to process user requests for food orders,
reservations, and inquiries.
● Order Management: Ability to manage food orders, including adding
items, modifying orders, and canceling orders.
● Reservation Management: Ability to handle table reservations, including
checking availability and confirming bookings.
● Menu Display: Displaying the menu items with descriptions and prices.
● Integration with GPT: Seamless integration with GPT for natural and
engaging conversations.
● Contextual Understanding: Ability to maintain context during the
conversation for a more natural interaction.
● Multi-Language Support: Support for multiple languages to cater to a
diverse user base.
● Payment Integration: Integration with payment gateways for secure and
convenient payments.
● Feedback Collection: Ability to collect user feedback to improve the
chatbot's performance over time.

Non Functional Requirements

● Performance: The chatbot should respond promptly to user queries and

handle multiple users simultaneously.
● Reliability: The chatbot should be available and reliable, with minimal
downtime.
● Scalability: Ability to handle a growing user base and increased traffic.
● Security: Ensuring that user data is secure and protected.
● Usability: The chatbot should be easy to use and navigate, even for users
unfamiliar with chatbots.
● Compatibility: Compatibility with various devices and platforms to
reach a wider audience.
● Maintainability: Easy to update and maintain to incorporate new
features and improvements.
● Privacy: Ensuring user privacy and compliance with data protection
regulations.

14
 3.2 Project Design and Architecture

Figure 1.1 Chatbot Flowchart

Dialogflow, a natural language understanding platform, processes user input

to understand intent and extract information. It then triggers appropriate
responses, which can include fetching data or performing actions from a
backend server. Dialogflow's integration with the backend enables dynamic
and personalized interactions, enhancing the chatbot's capabilities. The
backend server handles tasks such as processing orders, retrieving
information, and updating databases based on user requests. This integration
allows for a seamless and efficient chatbot experience for users.

15
 Figure 1.2 Dialogflow working

When a user enters or speaks an expression, Dialogflow matches the intent and
extracts parameters. It then sends a message to a webhook service containing
the intent, response, parameters, and actions. The webhook service performs
actions like API calls or database queries. Subsequently, the service sends a
response message back to Dialogflow, which forwards the final response to the
user for viewing or hearing.

3.2.1 Design of Security backend

Figure 1.3 User Authentication Flowchart

The backend of a login page securely verifies user credentials, typically a

password and a username or email. It retrieves the user's stored
information from a secure database using the provided username or email.
Passwords are stored in a hashed form, which is irreversible. During the
login process, the submitted password is hashed and compared to the
stored hashed password. Additional security measures like salting, which
adds random data to the password before hashing, are often used to
enhance security and prevent password attacks. This ensures user
authentication while protecting user accounts.
16
3.2.2 Steps & their Breakdown

1. In the DialogFlow console, create a new agent to represent the main

functionality of your chatbot. Also, enable the DialogFlow API for your GCP
project, facilitating seamless communication between your chatbot and the
API. This integration is crucial for the effective management of the chatbot's
natural language processing capabilities.

2. Create different objectives Within the Dialog Flow console, carefully outline
the different actions your chatbot should be able to handle. Additionally,
define entities that allow your chatbot to extract specific information from
user inputs, increasing its ability to understand and accurately respond to user
questions.

3. Spend time training your Dialog Flow agent by providing elaborate examples
of user inputs paired with desired responses. This training process enables the
underlying machine learning model to continuously improve its
understanding, refining the chatbot's conversational capabilities and ensuring
that it can effectively interpret and respond to user questions.

4. If your chatbot requires actions beyond simple responses, consider

implementing fulfillment webhooks. These server endpoints process user
requests and provide responses, allowing your chatbot to perform more
complex tasks or retrieve relevant data from external sources. Fulfillment
increases the functionality and usefulness of your chatbot.

5. Start rigorous testing in the Dialog flow console to simulate different user
interactions and evaluate your chatbot's responses. Extend testing to your
chosen integration platform to verify that the chatbot works seamlessly across
multiple environments and provides a consistent and reliable user experience.
Once you are satisfied with the performance of your chatbot, move from
testing to deployment. Let your chatbot live on your chosen integration
platform to interact with real users. This step marks the transition from
development to practical use, as users interact with the chatbot and receive
real-time responses.

6. Create a robust monitoring system that will continuously evaluate your

chatbot's performance through analytics and user feeDBack. Regularly
analyze user interactions and identify areas for improvement. Iterate on agent
intents, entities, and responses, incorporating user insights to improve the
chatbot's capabilities over time and ensure it remains dynamic and effective in
meeting user needs. Regular updates and continued user satisfaction.

17
3.3 Data Preparation

Data preparation is the process of preparing raw data so that it is suitable for further
processing and analysis. Key steps include collecting, cleaning, and labeling raw
data into a form suitable for machine learning (ML) algorithms and then exploring
and visualizing the data. Data preparation can take up to 80% of the time spent on
an ML project. Using specialized data preparation tools is important to optimize
this process.

3.3.1 NLU
Natural language know-how (NLU) is an important area of synthetic intelligence
that makes a specialty of the interaction among computer systems and human
language. With the use of natural language knowledge (NLU), computers can now
apprehend and interpret spoken language similarly to human beings. Many
strategies, together with deep learning, system learning, and natural language
processing, are used to gain this.
One of the important advantages of NLU is its ability to improve communication
among people and machines. NLU makes it feasible for machines to respond to
human language in a more intuitive and herbal manner, which enhances the
effectiveness and float of interactions with machines. Another gain is that NLU can
automate obligations that have been formerly finished via human beings. NLU, as
an example, automates customer service interactions in order that human people
can focus on more difficult obligations. As a result, organizations might save a
considerable amount of money and revel in increased purchaser pleasure.
Ultimately, NLU can be carried out to a variety of duties, which include sentiment
evaluation, chatbots, digital assistants, and textual content class. Because of the
increasing quantity of records generated daily, herbal language understanding
(NLU) is more important than ever in permitting computer systems to understand
and interpret human language in a meaningful way. This should notably exchange
the manner we interact with generation by making regular chores less difficult to do

18
3.3.2 NLP
"Natural Language Processing" (NLP) is a subfield of artificial intelligence that
researches the relationship among computer systems and human language. NLP is
the study of constructing fashions and algorithms that allow computer systems to
examine, understand, and write human language. This includes responsibilities like
language translation, text categorization, and sentiment analysis.
Among NLP's number one advantages is its capability to enhance conversation
among people and machines. NLP makes it viable for machines to reply to human
language in an extra intuitive and natural way, which will increase the effectiveness
and float of interactions with machines.
Another advantage of NLP is its capability to automate obligations that had been
previously completed with the aid of humans. For example, NLP automates
language translation in order that human translators can be conscious of greater
tough responsibilities. Companies might be capable of communicating more
efficiently and save a variety of cash.
Lastly, NLP is extensively carried out throughout several industries, together with
advertising and marketing, finance, and healthcare. NLP is more critical than ever
for giving computer systems the potential to apprehend and examine because of the
growing quantity of facts generated day by day, human language in a significant
manner. This could considerably exchange the manner we engage with generation
with the aid of making ordinary duties less difficult to complete.

3.3.3 NLP VS NLU

The synthetic intelligence subfields of Natural Language Understanding (NLU) and
Natural Language Processing (NLP) have a look at the interaction between
computer systems and human language. There are some vast differences among
them further to their shared characteristics.
Enabling computer systems to understand and understand human language in a way
that is just like that of human beings is the primary objective of herbal language
knowledge (NLU). This necessitates obligations like language translation, textual
content categorization, and sentiment evaluation is interested in the question of how
properly computer systems can realize the context wherein language is used. It
includes developing fashions and algorithms that allow computer systems to
understand and decode syntactic and semantic patterns in language.

19
 Figure 1.4 NLP w.r.t. NLP Diagram

3.3.4 Seq2Seq
Numerous herbal language processing responsibilities, such as device translation,
textual content summarization, and chatbot introduction, employ the powerful
series-to-series (Seq2Seq) deep studying structure. Main components of the Seq2Seq
structure are an encoder and a decoder. An encoder takes a set of input facts, like a
word in a language, and turns it into a hard and fast-duration vector illustration.
Subsequently, the decoder generates a series of output information, like a sentence in
an overseas language, the usage of this vector representation. With the Seq2Seq
structure, the model can learn how to map among the two sequences in a flexible and
adaptive way, which makes it specifically useful in situations where the lengths of the
input and output sequences fluctuate. Furthermore, by means of utilizing recurrent
neural networks (RNNs) in each the encoder and decoder, the version is able to
represent the temporal dependencies between the input and output sequences, making it
appropriate for tasks associated with herbal language processing. All things considered,
the Seq2Seq structure is an amazing instrument for a number of natural language
processing jobs, and builders and researchers in the discipline want it because of its
versatility and flexibility.

Figure 1.5 Selected competition

20
Performance Parameters
Sensitivity: ability of model to accommodate minor changes in dataset and input
by user. F1 score: Score generated based on precision and sensitivity of a model
FDR: false discovery rate, measure of incorrect output based on expected output.

3.3.5 Dataset
GPT 3.5 uses its very own device studying version to generate responses; this
version has been skilled on a large frame of literature from various domains and is
free of bias and offensive language.
The dataset used on this observation includes exchanges among a professional
psychologist and a psychological patient. The dataset changed into used to educate a
chatbot that can assist people with their mental health. Numerous questions on
mental fitness, such as ones about pressure, anxiety, and despair, are protected
within the dataset. The mental expert's responses are sponsored by way of their
know-how of psychology and they enjoy running in scientific settings. The series is
precise because it affords a plethora of facts on mental health issues and expert
remedies for them.
One of the primary blessings of this dataset is its usefulness in providing intellectual
health assistance. The dataset is a priceless useful resource for every person looking
for mental health care because the questions and answers are primarily based on
real- world occasions. The dataset is giant as it offers a couple of options for every
query and covers a lot of mental fitness conditions. This diversity will allow the
chatbot created with this dataset to correctly and as it should deal with a wide
variety of mental health problems. The chatbot's responses are ensured to be
subsidized via robust medical knowledge and competence with the aid of the replies
of a licensed psychological expert within the dataset. This dataset is usually helpful
for developing chatbots and other AI-based totally mental health aid systems

21
3.4 Implementation

Intents

Figure 1.6 Fallback Intent

The default fallback intent in Dialogflow serves as a safety net to handle user inputs
that do not match a certain intent. Fallback intent prevents communication
breakdowns when the chatbot encounters unclear or unexpected questions. A chatbot
allows you to respond politely with a standard message to ask for clarification or
guide the user to repeat their input. This provides a better user experience by
resolving unexpected interactions, improving chatbot adaptability, and maintaining a
smooth conversational flow.

Figure 1.7 Custom Payload

22
The intent of the menu in Dialogflow acts as a navigation guide, helping users
access specific functions or topics in the chatbot. By recognizing user requests
related to menu options, the intent is to direct users to relevant sections or
services, streamlining interactions. It provides a structured and user-friendly
experience that allows you to effectively explore the chatbot's capabilities. This
organized approach ensures that users can access the information or services they
want easily using the menu, improving usability and overall satisfaction.

Figure 1.8 Context

This Intent is responsible for adding items to the cart and to the mySQL database
and creating a record of all the items ordered and to be delivered, two parameters
are mandatory and work in it namely the quantity of food(@sys.number) and
food-item(@food-item) entity .

Figure

Figure 1.9 Intents

23
Entities

Figure 1.10 Entities

Entities in Dialogflow are critical components that improve chatbot understanding

and capture specific information from user input. This customizable element
identifies and categorizes details such as date, location, or product name, allowing
you to provide accurate answers. By setting a structure, developers enable chat
content to understand the user's context and provide more accurate and relevant
information. This dynamic feature greatly improves the chatbot's ability to
understand multiple questions, contributing to a more intelligent and context-aware
chat experience for users

24
Fulfillment

Figure 1.11 Webhook

Execution in Dialogflow means the execution of tasks outside of standard responses.

Accessed through webhooks, it allows a conversation to interact with external services,
databases, or perform custom actions. These powerful features enhance chatbot
capabilities and allow developers to integrate additional functionality. By improving
performance, chatbots can dynamically retrieve data, process transactions, or provide
users with a more personalized and comprehensive chat experience beyond predefined
responses.

Integration:
Dialogflow's Messenger integration makes it easy to seamlessly deploy chatbots across
messaging platforms. By connecting Dialogflow agents with platforms like Facebook
Messenger, developers extend the reach of their chatbots. Users can interact with the
chatbot directly in the messenger interface and receive real-time responses. This
integration simplifies user engagement and offers a familiar and convenient channel for
interactions. Leveraging Dialogflow capabilities through Messenger improves
accessibility and ensures a cohesive conversational experience for users across different
messaging platforms.

25
 Figure 1.12 API Integration

Backend Database

Figure 1.13 Menu table with item picking

26
Figure 1.14 Order item and total table

Figure 1.15 Order Tracking table

27
Code Snippets

Figure 2.1 Code snippet

FastAPI is a contemporary, brief (excessive-overall performance) net framework that

makes use of fashionable Python kind pointers to construct APIs with Python 3.7. It is
made to be brief to put in writing code and easy to apply. An incoming HTTP request is
represented by means of the FastAPI class Request. This is often used to system
incoming requests and extract information from them. A FastAPI class called JSON
Response is used to simulate an HTTP response with JSON content. It makes it simple
to return responses from your API endpoints in JSON format. It seems that the
document or module Database helper is being imported. Most possibly, that is a custom
module that gives help in interacting with a database. It may have features for starting
up database connections, strolling queries, and coping with database related operations.

Figure 3.11.2 Code snippet

Figure 2.2 Code snippet

28
The path is an installation to reply to HTTP POST requests at the foundation course
("/") with the aid of using the @app.Post("/") decorator. An asynchronous path handler
is what the handle request function does. It accepts an incoming HTTP request
represented by a Request object as a parameter. Using anticipate request.Json(), it
extracts the JSON statistics from the request body. It takes crucial records out of the
Dialogflow payload. The payload probably includes details about the user's intents,
parameters, output contexts, and session facts due to the fact Dialogflow is a platform
for herbal language understanding. Certain cause names are mapped to matching
handler functions (including add to order, remove from order, and so forth.) through the
dictionary intent_hand1er_dict. The extracted parameters and session ID are handed to
the applicable intent handler function, that's referred to as by way of the code based on
the received rationale.

Figure 2.3 Code snippet

First, the function calls DB helper.Get next order id() to get the subsequent order ID.
This means that the DB helper module carries a helper feature that, possibly from a
database, obtains the following available order ID. The gadgets in the order dictionary
are then iterated over, with every object representing a food item and its amount. The
meals object, amount, and order ID are inserted into the database for every item by way
of calling DBhelper.Insert order item. The feature returns -1 if the go back code (rcode)
is -1 after it's been checked. This can be a signal of an insertion manner mistake. The
feature inserts the order tracking repute into the database by means of calling
DBhelper.Insert order tracking after placing every person order object. The reputation
on this example is "in progress' '. At last next_order_id is used for further tracking on
identification.

29
 Figure 2.4 Code snippet

The parameters that the function calls for are session_id, that is a completely unique
identifier for the user session, and parameters, that are probably composed of the
person entered. From the parameters dictionary, it extracts the meals items and their
corresponding portions. It determines whether or not food items and quantities have
identical lengths. If no longer, it produces a reaction that suggests it doesn't realize
what the person has entered. When the lengths are identical, meals items and their
portions are paired to create a brand-new dictionary referred to as new food dict.

Figure 2.5 Code snippet

30
The characteristic starts off evolving with the aid of determining whether or not the
modern- day session's (session id) in-development order already exists. In the event
that it cannot locate the order, it could ask the user to place a new order in the response
that is back. In the event that an existing order is found, the items in the order are
removed in accordance with consumer enter. The meals gadgets which can be specific
are iterated thru and removed from the current_order dictionary. It continues a song of
objects that have been no longer within the order (no_such_items) and objects that have
been efficiently removed (removed_items). The function produces a response textual
content based totally on the elimination operation's results. The function returns a
JSON response containing the fulfillment text.

3.5 Key Challenges

Context Understanding
Achieving a thorough comprehension and reminiscence of context in a chatbot is vital
to imparting an effective and consumer-friendly enjoy, in particular in complex and
multi-flip interactions inclusive of food orders. In order for the chatbot to apprehend
context, state-of- the-art Natural Language Processing (NLP) capabilities ought to be
seamlessly included. This way decoding the underlying which means, rationale, and
context of the whole dialogue similarly to identifying particular consumer inputs.
Maintaining context becomes in particular tough but essential in terms of meal orders,
as conversations can take many exceptional twists and include distinctive functions
which include menu research, order customization, and address information. The
chatbot needs to be capable of understanding the person's preferences, follow the float
of the dialogue, and recall pertinent statistics from previous exchanges.

Natural Language Processing (NLP) Limitations

Overcoming NLP models' drawbacks, such as the possibility of misinterpreting unclear
user questions or the challenge of reliably extracting items from a variety of colloquial
languages Improving Natural Language Processing (NLP) models' shortcomings is
essential to improving chatbots' resilience and precision, especially when handling
complex user inquiries and a wide range of linguistic expressions. Mitigating the
possibility of erroneous user input interpretation is a noteworthy difficulty. Ambiguities
can be caused by homonyms, different sentence patterns, or colloquial language, which
makes it difficult for NLP models to correctly interpret the intended meaning.

31
It becomes essential to use cutting-edge NLP models with sophisticated contextual
awareness to get around this. Intricate contextual subtleties are well captured by models
such as BERT (Bidirectional Encoder Representations from Transformers) and GPT
(Generative Pre-trained Transformer).

Data Privacy and Security

In the design and functioning of a chatbot for food transport websites, protecting the
highest stage of safety and privacy of sensitive consumer data—consisting of payment
statistics and private statistics—is of essential significance. Strict adherence to criminal
rules and sturdy precautions to uphold personal confidence are essential while coping
with such sensitive data.
The use of cease-to-quit encryption is crucial for shielding sensitive personal facts.
This cryptographic protocol substantially lowers the threat of unauthorized access to at
some stage in statistics transfer by way of making sure that facts are safely transferred
and that simplest the legal recipient can decode it.

Integration with Backend System:

In the case of meals transport websites, achieving a clean interface among the chatbot
and backend structures is crucial to providing a unified and effective end-to-cease
experience. The procedure of integration entails coordinating the chatbot's features with
backend functions consisting of price gateways, stock control, and order processing.
First of all, adding order processing features guarantees that the chatbot can efficiently
receive, take care of, and validate purchase orders. This way creates an instantaneous
line of communique among the chatbot and the backend order control system so that
real-time updates at the status of orders, customization selections, and delivery data are
to be had.

Handling Multimodal Inputs

Addressing the developing need for non-textual interactions, including voice or image
recognition, and making sure the chatbot is capable of dealing with and reacting to
diverse inputs.

32
Dynamic Menu and Inventory Updates
In the context of food transport websites, a dynamic and adaptable chatbot must be able
to take care of real-time modifications to the menu and stock. This function allows a
streamlined and dependable person to revel in through making certain that consumers
acquire correct and present-day records approximately food gadgets, charges, and
promotions which can be accessible.
The backend systems in charge of handling the menu ought to be integrated with the
chatbot so that it is able to get actual-time updates every time something modifications.
In order to store menu information, synchronization with databases or Content
Management Systems (CMS) is required. The chatbot dynamically updates its database
to mirror changes in costs, descriptions, or the addition of recent items, giving users
access to the maximum latest information.

User Authentication and Authorization

Ensuring safe authorization and person identity verification during the chatbot ordering
technique on food delivery web sites requires the use of robust and intuitive
authentication techniques. This entails developing a clean, ultra-secure gadget that
protects person accounts even as simultaneously improving user revel in.
The chatbot should have safe authentication methods which will authenticate
customers. Multi-element authentication (MFA) techniques including email
confirmations, biometric authentication, and SMS verification codes are regularly
included on this. The implementation of multi-layered authentication improves
consumer account security through decreasing the opportunity of undesirable
admission to.

Generating Coherent and Human-like Response

Enhancing the generative version to generate responses which might be regular with
the tone and fashion of the brand, natural, and contextually suitable

33
 Chapter-04 System Development

4.1 Testing Strategy

User Testing
Software development manner: Testing man or woman elements, features, or modules
of the chatbot code one after the other is a vital step. Unit testing is a manner that
includes testing everything of the device one at a time to make certain features as
intended. Developers can locate and fix troubles early on by segmenting the code into
smaller, greater possible chunks.
Developers create test instances for precise functions or modules all through unit
testing, providing inputs and confirming that the outputs suit the meant consequences.
They can find and deal with insects, mistakes, or uncommon conduct in discrete
regions of the codebase thanks to this targeted method.

Integration Testing
Ensuring the general functionality and dependability of the device calls for verifying
the smooth interaction among the diverse components of the chatbot. This sort of
checking out, also known as integration testing, is focused on evaluating how different
components interact to produce the preferred end result.
Integration trying out for a chatbot might entail examining how essential components
work together, just like the database, the natural language processing (NLP) engine,
and any 1/3- birthday party APIs.

Functional Testing
The comprehensive system of testing the chatbot's universal functionality assesses how
properly the gadget performs in real situations. This sort of testing, additionally called
cease- to-give up trying out, evaluates how properly the chatbot can manipulate
consumer interactions from starting to stop. End-to-quit checking out incorporates
providing a lot of consumer inputs to the chatbot on the way to validate its natural
language processing (NLP) abilities. To assure that the chatbot efficiently translates
consumer motive, check instances have to encompass quite a number of query sorts,
including requests, commands, and questions.

34
User Scenario Testing

A vital component of comparing the chatbot's overall performance in actual-global

scenarios is trying it out with pre-made user eventualities. Using this approach,
builders can examine how properly the chatbot reacts to not unusual scenarios with the
aid of constructing a set of take a look at instances that mimic typical person
interactions. Often known as situation- based checking out or person scenario trying
out, this sort of testing seeks to discover feasible problems and assure an easy and
exciting consumer revel in. Developers can learn how well the chatbot plays in
scenarios that carefully resemble actual-global use through methodically trying it out
with pre-described consumer scenarios. This method aids within the early detection
and determination of viable issues in the course of the development procedure,
resulting within the improvement of a chatbot that offers customers with a reliable and
great enjoy at some stage in quite a number of usual interactions.

Usability Testing
A crucial first step in making certain a quality consumer revel in is assessing the
chatbot's usability and consumer interface. The interface layout, communique float,
person activate clarity, and response understandability are all carefully examined on
this assessment. Developers can layout a chatbot that is not simplest practical however
additionally clean to apply and apprehend by way of focusing on these elements.
Evaluating the chatbot's visual components, association, and basic layout aesthetics is
part of assessing the consumer interface. User engagement is substantially multiplied
by using simple, clean-to-use designs. Testers test for aesthetically alluring and
user-pleasant interface elements like shade schemes, font clarity, and prompt
placement.

Security Testing
It is crucial to verify the chatbot's safety so as to protect user information and
guarantee a reliable consumer experience. To find and cope with ability weaknesses
and threats, a radical assessment is needed. The testing technique covers a number of
safety-related subjects, including safeguarding conversation channels, shielding touchy
facts, and user inputs. Thoroughly validating person inputs is step one in safety testing.
The chatbot's potential to handle several inputs, inclusive of tries to inject malicious
code, SQL injection, or different popular input-based assaults, is classed by way of
testers. Making the chatbot properly verifies and sanitizes inputs reduces the
opportunity of protection lapses.

35
Load Testing
Evaluating the chatbot's potential to handle high volumes of visitors is essential to
guaranteeing its scalability and responsiveness, specifically in conditions wherein
there is an increase in consumer engagement. To discover viable bottlenecks and
improve the chatbot's basic overall performance, overall performance trying out
includes exposing it to an excessive quantity of concurrent users. In load checking out,
high person site visitors' situations are simulated to assess the chatbot's resilience.
Gradually increasing the number of concurrent users, testers preserve an eye fixed on
the machine's normal stability, reaction time, and aid utilization. This aids in locating
viable overall performance bottlenecks and the system's capacity limitations.

4.1 Test Cases and Outcomes

Test case 1:
Verify that the chatbot correctly identifies user intents related to ordering food tracking
orders, and other relevant actions
Input: Different food items and time history of recent ordered food as an input to the
function
Expected Outcome: The bot must be able to correctly identify the intent of user what
food item he/she want to order
Actual Outcome: The application correctly identifies the user intent and also gives
some suggestions from rating and new food items according to his/her taste from
history of ordering.

Test case 2:
To verify that the entities such as food items, quantities and delivery locations are
accurately extracted from user input.
Input: Quantity of the food the user provided and the items that the user wants as an
input and verify at the end of the order and providing the order number with the total
price.
Expected Outcome: Total number of items with the number of list of items if there is
same product then increasing its number and delivering the order number after placing
Actual Outcome: The application shows the accurate number of the list of items and
shows the accurate number of data and amount of price with accuracy.

36
Test case 3:
To Verify the Chatbot's capability to provide real-time tracking information
Input: The input will be the GPS data of the delivery boy with the distance between the
restaurant and user location.
Expected Outcome: Shows the accurate location of the delivery-man and expected time
to reach the destination.
Actual Outcome: The application shows the user the exact location and way the
delivery- man is approaching and estimated time to reach the final destination.

Test case 4:
To Verify that the user is able to modify the ordered item before placing the final order
Input: The saved data of the user before placing the order and list with proper number
of items Expected Outcome: Modified list must be shown to user and ask for final
placement of order
Actual Outcome: The application shows the modified list of items for the user and asks
the user to finalize the order if there is no further modification command.

Figure 3.1 Testing - Greeting & Place order

37
Figure 3.2 Testing - order.add intent & adding items

Figure 3.3 Testing - Menu intent

38
Figure 3.4 Testing the remove food-item intent

Figure 3.5 Testing the order.complete intent

39
 Figure 3.6 Testing the track order intent

Figure 3.7 ngrok localhost to server

Start your local server on the desired port, then open a terminal and navigate to the
NGrok directory. Run the command `ngrok http [port]`, replacing `[port]` with your
local server's port number. NGrok will generate a public URL that you can use to

40
access your local server from the internet.

Figure 3.8 Localhost Expess.js Server

This code sets up an Express.js server that utilizes the OpenAI API for text
generation. It initializes the OpenAI API client with an API key and defines a
function to generate text completions based on prompts. The server listens for
incoming requests, logging their paths and methods. For requests to the '/dialogflow'
endpoint, it handles 'input.unknown' actions from a Dialogflow webhook by
generating a text response using the OpenAI API and sending it back to Dialogflow.
Other actions result in a default message being returned.

41
 Chapter-05 Results and Evaluation

Results
The Generative AI chatbot has been successfully developed and tested. The chatbot has
been successfully able to identify all the intents i.e — fallback, welcome, menu, new
order, add order. Complete order, remove order and track order and reply back a human
like response with high accuracy, this integrated with a live interactive website that can
switch between dark and light mode and many other advanced CSS elements was our
final outcome.

Figure 4.1 Home Page UI

Integrated into live interactive websites, the Dialogflow chatbot provides a dynamic and
user- friendly interface. Upon entering the website, the user is greeted with a standard
welcome message initiated by the chatbot. This introductory message sets the tone for
giving a personal touch to user interactions. Using natural language processing,
chatbots answer user questions and guide the operation of various websites. This
integration improves the user experience, provides real-time help and information, and
makes the website more intuitive and attractive.

42
Figure 4.2 website with chatbot activated

Figure 4.3 Payment gateway

43
 Figure 5

Figure 4.4 Payment Successful

Figure 4.5 Menu and cart total

44
5.1 Comparison with existing Solutions
Our Dialogflow chatbot stands out among existing solutions because of its natural
language processing (NLP) capabilities and seamless integration with multiple
platforms. Unlike some alternatives, it harnesses the power of Google's Dialogflow,
providing an intuitive and user- friendly interface for effective user interaction. A set of
various predetermined goals, well- prepared organization, and effective training allows
for accurate understanding and memory of the context. Additional rich messaging
options increase user engagement, providing a visually pleasing and interactive
experience. In addition, the combination of implementation and webhook capabilities
allows chatbots to perform complex tasks and retrieve data in real- time, unlike
solutions with more limited functionality. Its adaptability, and regular updates based on
user feeDBack, position it as an advanced and evolving conversational agent, providing
a better user experience compared to the alternatives available in the NLP chatbot
landscape.

5.1.1 Understanding and Natural Language Processing (NLP):

Existing Solution: The generative AI chatbot has certain obstacles in the meanwhile,
mainly with regards to successfully knowledge and interpreting consumer input. This is
particularly sizable when answering uncertain questions or identifying problematic
linguistic nuances. These difficulties can also make it difficult for the chatbot to
respond with accuracy and contextual relevance, in order to negatively have an effect
on the person's enjoyment as a whole.
Improvements: Using state-of-the-art Natural Language Processing (NLP) algorithms
can significantly advance our potential to comprehend user input in a greater complex
way. Modern algorithms are extra adept at taking pictures of the subtleties of language,
allowing the chatbot to interpret uncertain requests and offer applicable answers.
Furthermore, it is crucial to integrate system learning models with a larger and extra
various schooling dataset. A large dataset can introduce the version to a greater
diversity of linguistic versions, facilitating its capability to modify to distinctive idioms
and conversation patterns.

5.1.2 Response Generation:

Existing Solution: The generative AI chatbot's cutting-edge country has a few
problems, in particular while it produces responses that might not be specific, clear, or
succinct in terms of the context. These issues may also bring about a decline in the
preferred satisfaction of the verbal exchange and in all likelihood in user discontent.

45
Improvements: It is critical to improve the response technology algorithms. This
includes going over the essential methods that generate responses again and refining
them to yield succinct, contextually correct responses. The chatbot's capability to
produce greater logical and pertinent responses can be advanced by honing those
algorithms, with the intention to enhance the general quality of user interactions.
Furthermore, making use of increasingly more complex language fashions is crucial for
raising the caliber of responses.

5.1.3 Handling Ambiguity and Uncertainty:

Existing Solution: When offered with uncertain personal inputs or scenarios outside of
its schooling statistics, the modern-day chatbot struggles and produces much less
beneficial or well-known responses. This restriction may additionally make it less
exciting for customers, especially in the event that they input unclear or non-anticipated
information.
Improvements: Several modifications may be made to the chatbot to deal with this
trouble and grow its adaptability. First of all, it is vital to put in force robust fallback
methods. By acting as a protection internet, fallback mechanisms assist the chatbot
reply intelligently to questions which can be uncertain or uncharted territory. In order to
make certain a more academic and encouraging interaction, those fallbacks can also
encompass predefined responses that direct users to clarify their inputs or provide more
details. Furthermore, it is vital to improve error message readability.

5.1.4 Scalability and Performance:

Existing Solution: Scalability issues with the contemporary chatbot gadget may want to
cause performance deterioration throughout durations of high visitors or height usage.
These problems underline the want for extra scalability as they are able to have an
effect on responsiveness and the overall person experience.
Improvements: A greater scalable and responsive gadget will result from streamlining
the chatbot's infrastructure, making use of effective resource control, and performing
performance checking out to discover and attach bottlenecks.

46
5.1.5 User Interface and Usability:
Existing Solution: There might be troubles with the chatbot's current person interface,
that could result in a much less-than-ideal user revel in. A multimodal approach is
usually recommended to enhance the overall usability and address these shortcomings.
Improvements: To make an interface that is each aesthetically attractive and clean to
apply, this entails honing the visual additives, layout, and standard aesthetics. Making
certain the layout is easy and clean to use can substantially enhance the person's
experience by means of facilitating user navigation and chatbot interplay. Another
important component of improvement is verbal exchange waft optimization. This
manner goes over the chatbot's logical collection of interactions to make sure the
communication makes sense and flows effortlessly.

5.1.6 Security and Policy:

Existing Solution: The chatbot device's cutting-edge nation shows possible security
flaws that could endanger user statistics and the system's basic integrity. A
comprehensive set of enhancements is essential to strengthen the safety posture and
protect in opposition to viable threats.
Improvements: The chatbot's safety posture may be strengthened and consumer data
included by means of wearing out sizable safety checking out, putting encryption
protocols into region, and making sure that it complies with industry requirements and
laws.

47
 Chapter-06 Conclusion and Future Scope

6.1 Conclusion
To sum up, creating and deploying a generative AI chatbot gives a modern way to
improve user revel in and expedite customer service. With the use of current gadget
getting to know and herbal language processing generation, this chatbot features a
digital assistant that may comprehend personal inquiries and offer human-like
responses.
Facilitating smooth and effective verbal exchange among customers and the food
delivery platform is the primary goal of the generative AI chatbot. Through using
modern language models, the chatbot is capable of interpreting a wide variety of user
input, together with one-of- a-kind language preferences, styles, and questions on the
meal shipping manner. This flexibility enables them to offer a customized and
easy-to-use enjoy that will increase purchaser happiness and loyalty.
A superior customer journey is facilitated by using the chatbot's capability to offer
actual- time assist, consisting of coping with rationalization, order monitoring, and
menu hints. The chatbot constantly learns and adapts so that it may provide accurate
and pertinent solutions to person requests through the years. This lessens the workload
for customer support groups and frees them up to pay attention to more complicated
problems, enhancing the personal experience standard.
Additionally, the generative AI chatbot helps the meal transport platform operate extra
correctly. The chatbot frees up human resources by handling a huge volume of
consumer interactions and automating repetitive operations, liberating the body of
workers' individuals to awareness of strategic jobs that call for a human touch. These
will increase in performance resulting in decreased expenses and higher use of sources
for the enterprise.
A scalable solution that may be grown and advanced upon over the years is the
generative AI chatbot, which adapts to converting personal expectations and
technological advancements. To upload new abilities, decorate language
comprehension, and preserve up with the modern- day developments in conversational
AI, normal enhancements and enhancements can be placed into place.
In conclusion, generative AI chatbots for food transport web sites are important for
updating purchaser interactions, increasing operational effectiveness, and keeping an
aggressive area within the speedy converting online meals services market. With its
deployment, a chief step has been taken towards providing purchasers with an extra
personalized, responsive, and intuitive revel in, for you to ultimately assist the food
transport platform prevail and ultimate.

48
6.2 Future Scope
The generative AI chatbot for food transport websites will remain evolved with an
emphasis on enhancing contextual focus, including multimodal interactions, and
utilizing state-of-the- art sentiment analysis for responses which are extra nuanced.
Machine getting to know techniques may be progressed, permitting the chatbot to alter
dynamically to convert user possibilities, dietary needs, and cultural quirks. An intuitive
and engaging ordering experience can be supplied with the aid of integration with
present day technology like speech recognition and augmented truth. Furthermore,
extending language aid and utilizing person comments to continuously train the model
could assure a person-targeted and the world over handy solution. Future plans include
developing a chatbot that is cleverer and more compassionate, highlighting its
significance in reshaping the marketplace for powerful and customized meal delivery
services.

49
References
1. Radford, A., Wu, J., Child, R., Luan, S., Sutskever, I., & Martens, I. (2019).
Language models are unsupervised feature learners. In: International
Conference on Learning Representations (ICLR).

2. Devlin, J., Chang, M.-W., Lee, K., Tabacoff, J., Venkatasubramanian, S.,
Vaswani, A., & Brundage, M. (2019). Bert: Pre-training of deep
bidirectional transformers for language understanding. In: North American
Association for Computational Linguistics (NAACL).

3. Liu, W., Zhou, S., Li, Z., & Liu, Z. (2020). UniLM: A unified language model
for natural language understanding and generation. In: Proceedings of the 58th
Annual Meeting of the Association for Computational Linguistics (ACL), pp.
4393-4404.

4. Yang, Z., Dai, Z., Yang, Y., Ye, J., Zhou, Z., & Liu, B. (2020). Ernie:
Enhanced representation through knowledge integration. In: Proceedings of
the 42nd ACM International Conference on Research in Computing and
Networking (SIGCOMM), pp. 4393-4404.

5. Rafael, C., Sharp, N., Narasimhan, E., Rush, A., & Su, S. (2020). XLNet:
Generalized autoregressive pre-training for language understanding. In:
Proceedings of the 34th International Conference on Machine Learning
(ICML),
pp. 7333-7343.

6. Gao, L., Yao, K., Chen, Y., Zhang, M., & Zhou, D. (2020). SimCSE: A simple
framework for semantic similarity learning using contrastive learning. In:
Proceedings of the 6th ACM Conference on Knowledge Graph and Semantic
Learning (KGSL), pp. 219-228.

7. Sun, X., Yang, Y., Dai, Z., Zhou, M., & Deng, L. (2021). ERNIE 2.0:
Enhanced neural representation with knowledge embedding. arXiv preprint
arXiv:2106.13796.

8. Liu, Z., Zhang, M., & Zhou, D. (2021). K-BERT: Knowledge-enhanced BERT
for improved question answering and natural language inference. arXiv
preprint arXiv:2106.00638.

9. Guu, K., Harutyunyan, H., Chen, T., Le, P., Duan, N., & Pascanu, R. (2021).

50
 Language models for dialog sum-marization: A survey. arXiv preprint.

10. Lee, H., Logeshwaran, L., Bengio, S., & Sutskever, I. (2021). Self-supervised
learning with contrastive language modeling. arXiv preprint arXiv:2107.12107.

11. Zhang, Z., Yang, S., Chen, Z., & Chen, M. (2021). Modeling dialogue history
for improved dialogue generation with recurrent self-attention networks. In:
Proceedings of the AAAI Conference on Artificial Intelligence (AAAI), pp.
11657- 11664.

12. Zhou, H., Yang, T., Huang, P., Zhang, Z., & Lan, W. (2021). UNITER:
Universal image- text representation learning. arXiv preprint
arXiv:2105.00049.

13. Lee, J., Kim, J., Kang, J., & Lee, H. (2022). Conversational AI: A survey.
arXiv preprint arXiv:2201.00944.

14. Zhang, Y., Zheng, S., Cheng, H., & Zhou, M. (2022). SimCSE++: A robust
framework for semantic similarity learning. arXiv preprint arXiv:2204.11892.

15. Wei, J., Zhang, S., Zhao, H., Liu, Z., & Zhou, M. (2022). P-tuning:
Empowering pre-trained language models for personalized dialogue
generation. arXiv preprint arXiv:2203.15828.

16. Li, X., Wu, Y., Zhang, T., Zhou, Y., & Zhang, Z. (2022). Dialogue
summarization with hierarchical attention networks. In:

17. Deng, L., Liu, B., Chen, Z., & Sun, X. (2022). ERNIE 3.0: Large-scale
knowledge enhanced pre-training for language understanding and generation.
arXiv preprint arXiv:2201.11122.

18. Chen, Z., Zhang, B., Zheng, S., & Zhou, M. (2022). SimCSE-B: A bi-encoder
architecture for robust semantic similarity learning. arXiv preprint
arXiv:2203.13058.

19. Yang, S., Zhang, Z., Chen, Z., & Chen, M. (2022). A unified framework for
multi- turn dialogue response generation with recurrent self-attention networks.
In: Proceedings of the 2022 Conference on Empirical Methods in Natural
Language Processing (EMNLP), pp. 1135- 1145.

51
20. Zhao, H., Wei, J., Zhang, S., Liu, Z., & Zhou, M. (2022). Personalized dialogue
generation with persona-aware pre-training. arXiv preprint arXiv:2204.04343.

21. Wang, P., Ji, M., Li, Z., Zhang, Y., Wang, L., & Zhang, Y. (2022). A survey on
knowledge- enhanced language models. arXiv preprint arXiv:2203.11747.

22. Liu, H., Chen, H., Zhou, M., & Zhang, Z. (2022). Enhancing dialogue
understanding with knowledge graph embeddings. In: Proceedings of the 50th
Annual Meeting of the Association for Computational Linguistics (ACL), pp.
754- 764.

23. R. Patel, et al., "Enhancing Chatbot Conversations using Generative AI: A

Comparative Study," in Proceedings of the IEEE International Conference on
Artificial Intelligence, 2023, pp. 112-120.

24. A. Kumar and B. Lee, "Ethical Implications of Generative AI in Chatbots: A

Case Study," in Proceedings of the IEEE International Conference on Ethics in
AI, 2024, pp. 230-238.

25. R. Patel, et al., "Enhancing Chatbot Conversations using Generative AI: A

Comparative Study," in Proceedings of the IEEE International Conference on
Artificial Intelligence, 2023, pp. 112-120.

26. J. Doe and A. Smith, "Advancements in Generative AI for Chatbot

Development," in IEEE Transactions on Artificial Intelligence, vol. 10, no. 3,
pp. 456-467, 2022.

52
Anirudh_Report
ORIGINALITY REPORT

9 %
SIMILARITY INDEX
9%
INTERNET SOURCES
2%
PUBLICATIONS
%
STUDENT PAPERS

PRIMARY SOURCES

1
ir.juit.ac.in:8080
Internet Source 4%
2
www.ir.juit.ac.in:8080
Internet Source 1%
3
www.coursehero.com
Internet Source 1%
4
aws.amazon.com
Internet Source 1%
5
www.geeksforgeeks.org
Internet Source <1 %
6
appquipo.com
Internet Source <1 %
7
lti.cmu.edu
Internet Source <1 %
8
www.aal-persona.org
Internet Source <1 %
9
www.ibm.com
Internet Source <1 %
10
ruor.uottawa.ca
Internet Source <1 %
11
www.scss.tcd.ie
Internet Source <1 %
12
www.ijraset.com
Internet Source <1 %
13
www.technavio.com
Internet Source <1 %
14
www.cmswire.com
Internet Source <1 %
15
edoc.pub
Internet Source <1 %
16
ijrpr.com
Internet Source <1 %
17
tudr.thapar.edu:8080
Internet Source <1 %
18
Viktar Atliha. "Improving image captioning
methods using machine learning
<1 %
approaches", Vilnius Gediminas Technical
University, 2023
Publication

19
dhruvdave.com
Internet Source <1 %
20
norr.numl.edu.pk
Internet Source <1 %
 JAYPEE UNIVERSITY OF INFORMATION TECHNOLOGY, WAKNAGHAT
PLAGIARISM VERIFICATION REPORT
Date: ………………………….
Type of Document (Tick): PhD Thesis M.Tech Dissertation/ Report B.Tech Project Report Paper

Name: ___________________________ __Department: _________________ Enrolment No _________

Contact No. ______________________________E-mail. ______________________________________
Name of the Supervisor: ________________________________________________________________
Title of the Thesis/Dissertation/Project Report/Paper (In Capital letters): ________________________
________________________________________________________________________________________________________

________________________________________________________________________________________________________
UNDERTAKING
I undertake that I am aware of the plagiarism related norms/ regulations, if I found guilty of any plagiarism and
copyright violations in the above thesis/report even after award of degree, the University reserves the rights to
withdraw/revoke my degree/report. Kindly allow me to avail Plagiarism verification report for the document
mentioned above.
Complete Thesis/Report Pages Detail:
 Total No. of Pages =
 Total No. of Preliminary pages =
 Total No. of pages accommodate bibliography/references =
(Signature of Student)
FOR DEPARTMENT USE
We have checked the thesis/report as per norms and found Similarity Index at ………………..(%). Therefore, we
are forwarding the complete thesis/report for final plagiarism check. The plagiarism verification report may be
handed over to the candidate.

(Signature of Guide/Supervisor) Signature of HOD

FOR LRC USE
The above document was scanned for plagiarism check. The outcome of the same is reported below:
Copy Received on Excluded Similarity Index Generated Plagiarism Report Details
(%) (Title, Abstract & Chapters)

Word Counts
 All Preliminary
Pages
Report Generated on  Bibliography/Ima Character Counts
ges/Quotes
Submission ID Total Pages Scanned
 14 Words String
File Size

Checked by
Name & Signature Librarian
……………………………………………………………………………………………………………………………………………………………………………

Please send your complete thesis/report in (PDF) with Title Page, Abstract and Chapters in (Word File)
through the supervisor at [email protected]