AI AND CHATBOT

Vedika Shinde Sakshi Gawai Srushti Auti Harish Ghodechor

ABSTRACT RELATEDWORK
Artificial Intelligence (AI) chatbots are intelligent systems designed The first chatbot ever was developed by MIT professor Joseph
to simulate human-like conversations through natural language Weizenbaum in the 1960’s. It was called ELIZA.
processing (NLP), machine learning (ML), and deep learning In the year 2009, a company called WeChat in China created a
techniques. more advanced Chatbot. Since its launch, WeChat has conquered
AI chatbots function as virtual assistants, capable of understanding the hearts of many users who demonstrate an unwavering loyalty
user queries, analyzing intent, and generating relevant responses to it. It is a highly thriving social media platform
in real time. Unlike rule-based systems, modern AI chatbots learn Early in 2016 we saw the intro of first wave of artificial data
continuously from interactions, thereby improving accuracy and technology in the design of chatbots. Social media platforms like
adaptability. Facebook enabled developers to build a chatbot for their
They are widely employed in customer service, healthcare, trademark or service so that customers could carry out some of
education, banking, and social media platforms, reducing their daily actions from inside their messaging platform.
operational costs while enhancing user experience. Furthermore, One of the early chatbots was ELIZA ,which was designed around a
AI chatbots support multi-language communication, context pattern matching approach. ELIZA was intended to sound like
awareness, and emotional intelligence to deliver human-like different persons, e.g. like a psychotherapist. Later, the
engagement. ELIZAinspired chatbot ALICE was introduced. ALICE was
implemented based on a scripting language called AIML1, which is
INTRODUCTION next to RiveScript which is currently among the most used
languages for chatbots. More recent chatbots, such as Rose,
Artificial Intelligence (AI) chatbots are intelligent systems designed combine chat scripts with question understanding to better mimic
to simulate human-like conversations through natural language domain experts [1]. Recent approaches have also considered
processing (NLP), machine learning (ML), and deep learning tackling the combination of RDF and chatbots, e.g., [2, 3, 9]. Due
techniques. to the lack of space, we refer the interested reader to [1, 7] for
AI chatbots function as virtual assistants, capable of understanding further references.
user queries, analyzing intent, and generating relevant responses
in real time. Unlike rule-based systems, modern AI chatbots learn
continuously from interactions, thereby improving accuracy and APPROACH
adaptability.
They are widely employed in customer service, healthcare, Our architecture is based on a modular approach to account for
education, banking, and social media platforms, reducing different types of user queries and responses, as can be seen in:
operational costs while enhancing user experience. Furthermore,
AI chatbots support multi-language communication, context User / Interface Layer
awareness, and emotional intelligence to deliver human-like This is the front-end where the user interacts with the chatbot .
engagement.
It could be a chat window on a website, a mobile app, or even
The main objective of AI chatbots is to automate communication voice-enabled assistants like Alexa or Google Assistance
and provide instant, efficient, and personalized support without
Example: A user types “Book me a train ticket from Pune to
human intervention. They are capable of handling simple queries
Mumbai.”
like FAQs as well as complex tasks such as booking tickets,
providing medical advice, or assisting in banking transactions. Input Proccesing Layer
With rapid advancements in AI technologies, modern chatbots are Takes the user’s raw input and prepares it for analysis.
moving beyond basic question-answering. They are now capable For text: performs tokenization, stop word removal, spelling correction.
of sentiment analysis, context retention, multilingual support, For voice: uses Speech-to-Text (STT) to convert speech into text.
and personalized recommendations, making them essential tools Example: Converts “Book me a train ticket from Pune to Mumbai”
in the digital era. into a cleaned text string.
Natural Language Processing (NLP) Layer
The brain of the chatbot. It tries to understand what the user
means.

1 [Link]
Components:
Intent Recognition → Detects what the user wants (e.g., book ticket).
Entity Recognition → Extracts key details (e.g., Pune, Mumbai, train).
Context Handling → Remembers past conversations if needed (e.g., return
journey info).
Example: Intent = Book Ticket, Entities = Source: Pune, Destination:
Mumbai.
Dialogue Management Layer
Decides how to respond based on the intent, entities, and context.
Components:
Policy Manager → Chooses next action (e.g., ask for date).
Response Generator → Creates the reply (predefined text,
dynamic text, or AI-generated).
Example: Decides to ask: “Sure! What date do you want to travel?”
Fig: Architecture Of Ai Chatbot
Knowledge Base / Backend Integration
Provides real data to the chatbot. LifeCycle Of Chatbot
Sources can include:
The life cycle of an AI chatbot involves systematic phases to
Databases (FAQ, customer info, booking system).
ensure efficient design, deployment, and continuous
APIs (railway booking API, weather, payment gateway). improvement. It can be summarized as follows:
Knowledge Graphs (semantic info).
Example: Connects to a railway API to check available trains 1. Requirement Analysis
between Pune and Mumbai. o Define objectives, target audience, and use
Natural Language Generation (NLG) Layer cases.
o Identify expected functionalities and success
Converts machine-readable response into human-like natural
metrics.
sentences.
2. Design Phase
Can use: o Develop conversational flow and interaction
Templates (e.g., “Your booking is confirmed for Train No. 1209 design.
from {source} to {destination} on {date}.”) o Select NLP techniques and frameworks.
AI models (e.g., GPT generating free-flow text). o Define chatbot personality, tone, and scope.
Example: Generates: “Your train ticket from Pune to Mumbai for 3. Development Phase
15th Sept is confirmed!” o Train NLP/NLU models for intent recognition
and entity extraction.
Output Layer o Integrate with databases, APIs, and external
Delivers the final response back to the user. systems.
Can be:Text → Shown in chat window. o Implement dialogue management and
response generation.
Voice → Converted via Text-to-Speech (TTS).
4. Testing Phase
Multimedia → Buttons, images, booking confirmations. o Perform functional, usability, and
Example: User sees a message with a booking ID + option to performance testing.
download ticket. o Evaluate accuracy of intent recognition and
quality of responses.
Flow Recap (Diagram Explanation in Words):
5. Deployment Phase
User enters query → (Interface Layer). o Launch chatbot across multiple platforms
Query processed → (Input Processing Layer). (web, mobile, social media).
Chatbot understands meaning → (NLP Layer). o Ensure scalability, security, and compliance.
Decides best response → (Dialogue Management). 6. User Interaction & Feedback
Fetches real data if needed → (Knowledge Base/API). o Monitor real-time conversations and collect
feedback.
Converts response into natural sentence → (NLG Layer).
o Analyze user satisfaction and chatbot
Delivers back to user → (Output Layer). efficiency.
7. Maintenance & Continuous Improvement
o Update intents, responses, and knowledge
base regularly.
o Retrain AI models for better accuracy and
adaptability.
o Enhance features to align with evolving use
Cassification Of A  Conversational AI Chatbots → Advanced, can hold
human-like conversations, learn continuously.

Conclusion
The main objectives of the project were to develop an algorithm
that will be used to identify answers related to user submitted
questions. To develop a database were all the related data will be
stored and to develop a web interface. The web interface
developed had two parts, one for simple users and one for the
administrator. A background research took place, which included
an overview of the conversation procedure and any relevant chat
bots available. A database was developed, which stores
information about questions, answers, keywords, logs and
feedback messages. A usable system was designed, developed
Classification Of Chatbot and deployed to the web server on two occasions. An evaluation
took place from data collected by potential students of the
Based on Functionality University. Also after received feedback from the first
deployment, extra requirements were
Rule-based ChatbotsFollow predefined rules and [Link] introduced and implemented. AI chatbots have emerged as
with keywords and fixed [Link]: FAQ bots. powerful tools that transform the way humans interact with
technology. By combining natural language processing, machine
learning, and conversational interfaces, they provide quick,
 AI-powered Chatbots efficient, and personalized communication. They are widely used
o Use NLP (Natural Language Processing) and in industries like education, healthcare, e-commerce, and
ML (Machine Learning). customer service, offering 24/7 support and reducing operational
o Understand context, intent, and can learn [Link] chatbots cannot fully replace human intelligence,
over time. they complement human efforts by handling repetitive tasks,
o Example: Siri, Alexa, Google Assistant. improving user experience, and enabling businesses to scale. With
continuous advancements in AI, future chatbots will become
Based on Communication Style more human-like, emotionally intelligent, and context-aware,
making them an integral part of digital communication.
Text-based Chatbots → Interact using only text. (e.g., website
chat support) Reference
Voice-based Chatbots → Use speech recognition + text-to-
speech. (e.g., Alexa, Google Assistant)
[1] Sameera A Abdul-Kader and John Woods. 2015. Survey on chatbot design
techniques in speech conversation systems. Int. J. Adv. Comput. Sci. Appl.
Multimodal Chatbots → Support text, voice, images, and buttons. (2015).
[2] H. Al-Zubaide and A. Issa. 2011. OntBot: Ontology based chatbot. In
International Symposium on Innovations in Information and Communications
3. Based on Use Case / Application Technology.
[3] A. Augello, G. Pilato, G. Vassallo, and S. Gaglio. 2009. A Semantic Layer on Semi-
Structured Data Sources for Intuitive Chatbots. In 2009 International
Customer Support Bots – Handle queries, complaints, FAQs. Conference on Complex, Intelligent and Software Intensive Systems. 760–765.
[4] Andreas Both, Dennis Diefenbach, Kuldeep Singh, Saeedeh Shekarpour, Didier
Cherix, and Christoph Lange. 2016. Qanary - A Methodology for
E-commerce Bots – Help with shopping, recommendations,
VocabularyDriven Open Question Answering Systems. In ESWC. 625–641.
orders. [5] Timofey Ermilov, Diego Moussallem, Ricardo Usbeck, and Axel-Cyrille Ngonga
Ngomo. 2017. GENESIS: a generic RDF data access interface. In Proceedings of
the International Conference on Web Intelligence. 125–131.
Healthcare Bots – Provide symptom checking, appointments. [6] Jens Lehmann, Robert Isele, Max Jakob, Anja Jentzsch, Dimitris Kontokostas,
Pablo N Mendes, Sebastian Hellmann, Mohamed Morsey, Patrick Van Kleef,
Sören Auer, et al. 2015. DBpedia–a large-scale, multilingual knowledge base
Educational Bots – Tutoring, quizzes, student support.
extracted from Wikipedia. Semantic Web 6, 2 (2015), 167–195.
[7] Michael McTear, Zoraida Callejas, and David Griol. 2016. Conversational
Personal Assistant Bots – Schedule meetings, reminders. Interfaces: Past and Present. In The Conversational Interface. Springer, 51–72.
[8] Pablo N. Mendes, Max Jakob, Andrés García-Silva, and Christian Bizer. 2011.
DBpedia spotlight: shedding light on the web of documents. In Proceedings the
. Based on Level of Intelligence 7th International Conference on Semantic Systems, I-SEMANTICS. 1 – 8.
[9] Mon-Tin Tzou, Chun-Hung Lu, Chin-Chien Wang, Cheng-Wei Lee, and Wen-Lian
Hsu. [n. d.]. Extending knowledge of AIML by using RDF. ([n. d.]).
Simple Chatbots → Limited to fixed Q&A. [10] Richard S Wallace. 2009. The anatomy of ALICE. Parsing the Turing Test (2009)
, 181–[Link] Weizenbaum. 1966. ELIZA—a computer program for the
study of natural language communication between man and machine.
Smart Chatbots → Understand context, sentiment, and give Commun. ACM (1966).
dynamic responses.
Joseph Weizenbaum. 1966. ELIZA—a computer program for the
study of natural language communication between man and
machine. Commun. ACM (1966).