A Project Report

on
DocTalk
(a healthcare assistant)

submitted in partial fulfilment of the requirement for the award of the degree

of

Bachelor of Technology
in Computer Science Engineering and Information Technology
by
Harshit 21CSU036

Under the supervision of

Dr. Vandana

Department of CSE & IT

The NorthCap University, Gurugram
May 2024
 CERTIFICATE

This is to certify that the Project Synopsis entitled, “DocTalk” submitted by “Harshit” to The

NorthCap University, Gurugram, India, is a record of bonafide synopsis work carried out by

them under my supervision and guidance and is worthy of consideration for the partial fulfilment

of the degree of Bachelor of Technology in Computer Science and Engineering of the

University.

Dr. Vandana

Date: ………………
 ACKNOWLEDGEMENT

No project is ever the result of a single individual's effort. It reflects the

collective support, guidance, and contributions of many people who,
directly or indirectly, played a role in its successful completion.
Before delving into the details of this project, we would like to take a
moment to express our sincere gratitude to those who stood by us
throughout this journey. Their encouragement, insight, and timely help
made this undertaking possible.
First and foremost, we are deeply grateful to Dr. Vandana for her
invaluable guidance, constant support, constructive feedback, and
encouragement throughout the course of this project. Despite her busy
schedule, she always made time to guide us, for which we are truly
thankful.
We would also like to extend our heartfelt thanks to our friends, who
assisted us with dedication and enthusiasm. Their input, discussions, and
collaborative spirit helped us overcome many challenges we encountered
along the way.
Lastly, we express our sincere appreciation to our respected Head of
Department, Dr. Rita Chhikara, for her unwavering support and for
constantly motivating us to strive for excellence. Her positive outlook and
words of encouragement inspired us to stay committed and focused
throughout the project.
This project would not have been possible without the collective effort of
all these wonderful individuals, and we are truly grateful to each one of
them.

Harshit Beniwal 21CSU036

 INDEX

1. Abstract

2. Introduction (description of broad topic)

a. Background
b. Feasibility Study

3. Study of Existing Solution (implementation-based project)/Literature Survey

(research based)
4. Comparison with existing software solutions (implementation
based)/Literature (research based)
5. Gap Analysis
6. Problem Statement
7. Objectives
8. Outcomes
9. Gantt Chart
10.Responsibility Chart
11.References
12.Annexure I: Front page of plagiarism report by guide
13.Annexure II: Screenshots of Faculty mail / comments from guide
 LIST OF TABLES
1. TABLE 1: The table below compares and contrasts the top existing
solutions
2. TABLE 2: Available platforms
3. TABLE 3: Gaps in existing solutions
4. TABLE 4: Gantt Chart
5. TABLE 5: Responsibility Chart
 ABSTRACT

With the rapid advancement of artificial intelligence, healthcare assistant

chatbots have emerged as valuable tools for enhancing patient care by
automating responses to common medical queries.
This project focuses on the development of an intelligent chatbot designed to
interact with users, understand their symptoms, and suggest possible remedies or
next steps based on predefined medical knowledge and conditions.
By integrating Natural Language Processing (NLP) and machine learning
techniques, the chatbot aims to offer a user-friendly, accessible platform for
preliminary medical support.
It can assist users in identifying potential illnesses, provide information about
symptoms and medications, and advise when it is necessary to seek professional
medical consultation.
Ultimately, the goal is to create an initial layer of medical guidance that can either
support hospital systems or function independently as a web-based healthcare
assistant, offering quick, informative, and responsible first-level medical
interaction.
 INTRODUCTION

As Artificial Intelligence (AI) continues to evolve at a rapid pace, its applications

in healthcare have shown promising potential to address some of the most
pressing challenges faced by the sector. Among these innovations, healthcare
assistant chatbots have emerged as effective tools for managing patient
volumes, providing timely information, and reducing the burden on healthcare
professionals. These intelligent systems aim to simulate conversations similar
to those between patients and healthcare providers, offering guidance on non-
emergency medical concerns and helping individuals decide whether they
need to consult a doctor in person.
Healthcare chatbots operate by interpreting user input related to symptoms
and returning relevant health information or recommendations. By leveraging
natural language processing and AI-driven algorithms, they can deliver
accessible, real-time support that empowers users to make informed health
decisions.
The use of such technology gained significant momentum during the COVID-19
pandemic, when healthcare systems across the globe were overwhelmed.
Governments and medical institutions turned to AI solutions, including
chatbots, to meet the surge in demand for remote consultations and self-
assessment tools. Beyond pandemic response, these tools offer ongoing value
by enhancing healthcare accessibility—especially in remote or underserved
areas—and by supporting clinicians with automated, routine interactions.
This project focuses on the development of a Health Assistant Chatbot that
can serve as a first point of contact for users seeking medical advice. It is
designed to provide symptom-based recommendations and help streamline
the decision-making process for patients, making healthcare support more
efficient, reliable, and user-friendly.
a. Background

The global healthcare ecosystem is under mounting strain due to a

combination of growing patient populations, increasingly complex medical
conditions, and limitations in healthcare infrastructure and personnel. These
challenges are not confined to any one region—they are felt worldwide,
especially in underserved communities and resource-constrained
environments. The imbalance between the rising demand for medical
attention and the available healthcare workforce has created significant gaps
in timely and effective care delivery.
In response, Artificial Intelligence (AI) has begun to reshape how healthcare
services are delivered, offering tools that can support clinicians, improve
patient engagement, and reduce operational burdens. While AI-powered
technologies have already gained considerable traction in sectors like
customer service and finance, their application in healthcare is now gaining
momentum, particularly through conversational AI tools like chatbots.
Healthcare assistant chatbots are designed to automate various tasks that
would otherwise require human intervention. These range from answering
frequently asked questions and performing basic symptom triage to
connecting patients with appropriate healthcare providers. By using Natural
Language Processing (NLP), these bots can understand and respond to user
queries in a manner that closely mimics human interaction. This makes them
especially valuable for providing instant, around-the-clock support—even in
areas where medical professionals are not readily available.
One of the most significant advantages of healthcare chatbots is their cost-
effectiveness and scalability. Unlike human-operated services, they can
handle thousands of interactions simultaneously at minimal cost, making them
ideal for supporting large populations. Additionally, their use can be tailored to
address public health gaps by offering guidance for common medical issues,
managing follow-up care, and promoting early intervention—all of which
contribute to easing the burden on overworked health systems.
This project builds on these advancements by developing a Healthcare
Assistant Chatbot that focuses on common medical conditions. It is
programmed with a curated database of symptom-response mappings and
provides preliminary guidance based on user input. Moreover, the chatbot can
suggest when professional consultation is necessary, acting as a first line of
support for individuals seeking quick and reliable health information.
b. Feasibility study

● Technical Feasibility:

1. Technology Used:
a. Python: Backend logic and machine learning model
implementation.
b. Tkinter: For creating the graphical user interface (GUI).
c. Scikit-learn: For building and training the decision tree model.
d. Pandas and NumPy: For handling and analyzing the data.
e. Matplotlib: For visualizing results, if applicable.

2. Key Issues:
a. Optimizing the decision tree classifier for accurate predictions.
b. Ensuring the GUI is user-friendly and responsive.
c. Integrating datasets (Training.csv and Testing.csv)
effectively.

3. Project Timeline: Estimated at 3-5 months for development, testing, and

deployment.

● Financial Feasibility

1. Overall expenses: Minimal costs, as it primarily involves open-

source tools and frameworks.
2. Financing: External funding is not required.

3. Revenue: Could be monetized through partnerships with

healthcare providers or advertisements once deployed.
 ● Market Feasibility

Demand:

 Increasing demand for AI-driven healthcare solutions, particularly chatbots

capable of providing initial diagnoses and recommendations

 A growing focus on accessible healthcare through digital platforms.

Differentiation:

 The chatbot’s ability to suggest diagnoses and recommend specialists

distinguishes it from generic healthcare chatbots.

Competition:

 While some existing solutions like WebMD or symptom-checking apps

exist, they are often not specialized for localized or easily accessible
recommendations.

Potential Users:

 Individuals seeking quick and accessible medical advice.

 Medical practitioners interested in integrating chatbots for triaging patients .
 STUDY OF EXISTING SOLUTION

In recent years, a variety of AI-powered platforms have emerged in the digital

healthcare space, aiming to assist users with symptom checking, health guidance,
and basic medical information. These tools represent meaningful strides in the
integration of artificial intelligence into everyday health management. However,
despite their innovation, most existing solutions still fall short in delivering a
holistic, real-time, and truly personalized healthcare experience.
Below are some of the most well-known healthcare assistant tools that have shaped
the current AI landscape, along with a critical evaluation of their features and
limitations:

1. WebMD Symptom Checker

 Features:
WebMD offers a basic symptom-checking tool that allows users to enter
symptoms and receive a list of potential conditions that might be causing
them. The interface is straightforward and accessible via the web and mobile
app.
 Limitations:
WebMD’s system relies primarily on static symptom-condition mappings
rather than dynamic AI learning. It does not adapt over time or learn from
user data, which limits its diagnostic depth. Additionally, it lacks interactive
capabilities—users do not receive tailored advice or follow-up questions,
and there's no real-time engagement with a healthcare provider or specialist.

2. Ada Health
 Features:
Ada Health employs a more sophisticated AI engine that mimics medical
decision trees, asking users a series of detailed questions based on their
symptoms. The tool offers symptom assessments with a high level of
accuracy, backed by a medically-reviewed knowledge base.
 Limitations:
While Ada offers strong diagnostic capabilities, a significant portion of its
advanced features is locked behind a subscription model. The tool also does
not offer a direct path to healthcare professionals or provide specialist
 referrals, which limits its usability for users seeking next-step guidance
beyond self-assessment.

3. Healthily (formerly Your.MD)

 Features:
Healthily provides a user-friendly chatbot that supports basic health queries
and symptom checking. It emphasizes self-care and encourages users to
make informed decisions for managing everyday health issues. It also
includes general wellness tools, such as health trackers and condition
libraries.
 Limitations:
Although Healthily is helpful for minor health concerns, it does not support
real-time decision-making or personalized medical recommendations. The
chatbot operates within a narrow scope and does not escalate care to human
doctors or specialists, which can limit its usefulness in more serious or
complex situations.

Insights and Observations

These platforms have played a foundational role in demonstrating how AI can
support healthcare delivery. However, their shortcomings underscore the need for
solutions that are:
 More Accessible: Many platforms offer only partial functionality for free
users.
 More Personalized: Few offer localized insights or adapt to user-specific
conditions.
 More Actionable: There is often a gap in connecting users to actual
healthcare professionals or offering next steps.
 More Adaptive: Most rely on static logic and do not integrate machine
learning models that evolve with time or user feedback.

How Our Project Fills These Gaps

In response to these limitations, our project introduces a Healthcare Assistant
Chatbot that combines affordability, real-time interaction, and intelligent decision-
making. Unlike most existing tools, our chatbot:
  Uses a Decision Tree Classifier for dynamic, pattern-based analysis of
symptoms.
 Provides specialist recommendations, bridging the gap between self-
assessment and professional care.
 Is completely free, eliminating access barriers common in freemium
models.
 Incorporates local health data to deliver contextually relevant responses,
especially for underrepresented populations.
By learning from the strengths and weaknesses of existing platforms, our solution
aims to offer a more complete, user-centered healthcare assistant that is both
practical and inclusive.
 Comparison with Existing Solutions

The landscape of AI-powered healthcare chatbots is evolving rapidly, with

numerous platforms offering symptom-checking and basic medical guidance.
However, most of these tools face critical limitations in accuracy, personalization,
accessibility, and usability. Our proposed healthcare assistant chatbot aims to fill
these gaps with a more intelligent, user-focused design. A detailed comparison is
presented below:

1. Accuracy and Diagnostic Capability

Many popular platforms such as WebMD or Healthily rely on static, rule-based systems. These
systems operate on pre-defined logic trees that fail to adapt to the diverse and often complex
nature of real-world medical symptoms. As a result, their responses can feel generic, and
diagnostic accuracy may be limited when dealing with overlapping or rare conditions.
In contrast, our chatbot integrates a Decision Tree Classifier, a machine learning algorithm that
learns from real-world data. This allows the system to uncover intricate relationships between
symptoms and conditions. Rather than relying on rigid logic, our model dynamically adjusts its
predictions based on the patterns found in the training data. The result is a more adaptive, data-
driven chatbot capable of delivering precise responses tailored to the user's symptoms.

2. Actionable Specialist Recommendations

A recurring shortfall in existing chatbots is the absence of actionable advice. While they might
suggest potential conditions, they often stop short of guiding the user on what to do next. This
leaves individuals uncertain about whether to seek professional care or which specialist to
consult.
Our chatbot fills this gap by offering clear, actionable recommendations, including referrals to
relevant types of medical professionals such as dermatologists, general physicians, or
cardiologists based on the user’s symptom profile. This feature empowers users with the
knowledge needed to make informed healthcare decisions, transforming passive symptom-
checking into proactive health management.
3. Accessibility and Cost-effectiveness
High-quality medical tools are often behind paywalls, especially in apps like Ada Health or
advanced WebMD subscriptions. This limits access to individuals in low-income regions or
those unable to pay for premium services, thereby widening the healthcare accessibility gap.
Our solution is completely free to use, removing economic barriers and making medical
guidance more inclusive. Whether someone is living in a metropolitan area or a remote village,
they can access the chatbot without subscriptions or hidden costs. This commitment to open
access is crucial in democratizing healthcare technology.

4. Customization and Local Relevance

Most global platforms cater to a general audience, which means they often overlook region-
specific health issues or cultural nuances in symptom descriptions. This can lead to irrelevant or
inaccurate suggestions for users in different parts of the world.
To address this, our chatbot is trained on datasets curated with a focus on local and commonly
occurring health conditions. This local adaptation ensures that the suggestions provided are not
only accurate but also relevant to the community the user belongs to. From region-specific
illnesses to familiar terminology, the chatbot becomes more relatable and effective for its
intended audience.

Conclusion
In summary, while existing healthcare chatbots provide a valuable foundation, they
fall short in several key areas. Our project introduces a refined approach that
prioritizes machine learning-driven accuracy, real-time specialist
recommendations, cost-free access, and localized customization. These
enhancements make our chatbot not just a tool—but a practical, supportive digital
health companion for users at all levels of digital literacy and income.

The table below compares and contrasts “Vani Sanket” with the top existing solutions:
 Tool Accuracy Languages Specialist Pricing Special Features
Supported Recomme
ndations

WebMD medium English No Free Basic symptom

checker, generic
advice

Ada Health High English, No Freemiu AI-based

limited m symptom
others analysis, paid
advanced features

Healthily high English No Freemiu Self-care advice,

m symptom analysis

DocTalk high English Yes Free Symptom-based

diagnosis,
decision tree
model, doctor
recommendations

GAP ANALYSIS
A thorough analysis of existing AI-powered healthcare chatbot platforms reveals
several fundamental gaps that hinder their ability to serve users effectively,
especially in diverse and resource-constrained settings. These shortcomings point
to the urgent need for a more adaptive, intelligent, and user-centric healthcare
solution. The key gaps identified are as follows:

1. Limited Diagnostic Accuracy and Flexibility

Most existing platforms rely on predefined, rule-based systems that follow scripted logic to
match symptoms with possible conditions. While this approach can handle basic queries, it fails
when users present with multiple symptoms or complex health histories. These static models
cannot adapt to new medical data or learn from past interactions, leading to limited diagnostic
precision. In contrast, machine learning models, such as decision trees, offer a dynamic
solution by learning from real-world data and continuously improving their predictions over time
—something sorely missing in many current solutions.

2. Absence of Specialist Guidance

A major limitation across many tools is the lack of next-step guidance. Most chatbots stop at
listing potential ailments without suggesting whether the condition warrants medical attention or
which type of healthcare professional should be consulted. This leaves users in a state of
uncertainty, unsure whether they need to seek urgent care, consult a specialist, or simply
monitor symptoms at home. Without actionable follow-up, the tool's value in decision-making
remains incomplete.

3. Accessibility and Financial Barriers

Popular platforms often operate on a freemium or tiered subscription model, where basic
features are offered for free, but critical diagnostic insights, live consultations, or advanced
assessments are locked behind paywalls. This monetization strategy disproportionately affects
individuals in low-income or underserved communities, creating a digital divide where only
those who can afford premium access receive quality care. For a solution to be truly inclusive,
healthcare guidance should not come at a cost—especially when dealing with potentially
serious medical concerns.

4. Poor User Experience and Usability Issues

Many users—especially those who are elderly, unfamiliar with technology, or dealing with stress
due to illness—struggle to navigate complex interfaces. Current platforms often feature
cluttered designs, confusing navigation, or slow response times, which can discourage users
from continuing. Moreover, the lack of multilingual support in many chatbots limits their
adoption in regions where English is not the first language, thereby reducing accessibility for
vast populations.

5. Insufficient Localization and Cultural Relevance

Most existing chatbot solutions are designed with a global audience in mind and do not reflect
localized health trends, cultural sensitivities, or region-specific diseases. As a result, their
recommendations may be irrelevant or inappropriate for users from different parts of the world.
A health assistant intended for India, for instance, should be capable of addressing conditions
prevalent in the region (e.g., dengue, chikungunya) and understanding local health practices. The
absence of such localization leads to reduced trust and effectiveness in real-world applications.

Conclusion
The current generation of healthcare chatbots has laid the foundation for AI-driven
medical assistance, but critical gaps remain in terms of diagnostic intelligence,
guidance, inclusivity, and cultural adaptability. These findings strongly justify the
need for a more robust, affordable, personalized, and locally aware healthcare
assistant chatbot—one that goes beyond symptom-checking to truly empower
users with accurate advice and practical next steps.

How DocTalk Addresses These Gaps

To bridge the limitations found in existing AI healthcare tools, DocTalk has been
strategically designed with a user-centric approach, integrating modern machine
learning techniques and usability-focused design. The following features
collectively empower DocTalk to provide a more reliable, accessible, and
regionally relevant healthcare assistant experience:

1. Enhanced Diagnostic Accuracy through Machine Learning

Unlike traditional rule-based systems, DocTalk leverages a Decision Tree Classifier, a
supervised machine learning algorithm capable of learning from historical medical data. This
allows the chatbot to:
 Identify patterns and correlations between symptoms and conditions.
 Deliver more personalized and data-driven assessments.
 Improve over time as more data becomes available or retrained.
This approach significantly enhances the diagnostic precision of the chatbot, reducing the risk of
generic or misleading outputs and offering a more confident first layer of support for users.

2. Actionable Specialist Recommendations

A critical improvement over many existing platforms is DocTalk’s ability to guide users toward
the right type of medical specialist based on their symptoms. This feature:
 Reduces user confusion about next steps.
 Helps avoid misdirected consultations.
 Promotes timely and targeted medical attention.
For example, if a user reports persistent chest pain, DocTalk may suggest consulting a
cardiologist rather than offering only condition names. This bridges the gap between symptom
checking and practical medical guidance.

3. Free and Open Access

Recognizing the need for inclusive and equitable healthcare support, DocTalk is offered
completely free of charge. This decision addresses several key issues:
 Removes the paywall barrier present in many commercial platforms.
 Enables users from underserved or low-income communities to access reliable health
information.
 Ensures critical healthcare resources are not limited to those who can afford premium
subscriptions.
This open-access model reinforces the mission of making digital health tools universally
accessible.
4. Simple and Intuitive User Interface
DocTalk is built with Tkinter, a lightweight and accessible GUI library in Python. The interface
is designed to:
 Be clean, minimalistic, and clutter-free.
 Support ease of use for individuals with low technical proficiency.
 Offer a smooth conversational experience that mimics natural dialogue.
By lowering the technological barrier, DocTalk becomes accessible to a wider demographic,
including elderly users or those unfamiliar with digital tools.

5. Regional Relevance through Localized Training Data

One of the standout features of DocTalk is its localization strategy. The chatbot is trained on
datasets that reflect region-specific health concerns, enabling it to:
 Deliver context-aware responses tailored to local healthcare patterns.
 Identify symptoms that are more common in specific geographic areas.
 Improve user trust and satisfaction by offering advice that feels more relevant and
applicable.
This localization ensures that DocTalk remains effective not just globally, but within the cultural
and epidemiological context of its users.

Summary
By combining intelligent diagnostics, practical guidance, free accessibility, and
thoughtful interface design, DocTalk successfully overcomes the major limitations
of existing healthcare chatbots. It represents a meaningful step forward in
democratizing healthcare support and empowering users to take proactive,
informed steps in their healthcare journey.
 PROBLEM STATEMENT

In an era of rapid technological advancement, artificial intelligence (AI) is

increasingly being leveraged to address critical gaps in healthcare. AI-powered
chatbots and symptom checkers have emerged as scalable tools designed to support
users with medical queries, self-assessments, and health decision-making.
However, despite their growing presence and potential, these tools still face a
range of significant challenges that hinder their effectiveness, inclusivity, and real-
world impact.
The primary issues that continue to affect current AI-driven healthcare platforms
are outlined below:

1. Inaccuracy in Symptom Assessment

Most widely-used symptom checker tools rely on static, rule-based systems. These systems
follow predefined logic paths and are unable to adapt to the diverse and often complex nature of
human health. Consequently, users frequently receive generalized or overly broad diagnostic
suggestions that fail to capture the nuances of individual symptoms or multiple co-occurring
conditions. This compromises the trustworthiness of these tools and can potentially delay
appropriate care.

2. Lack of Specialist Recommendations

Although existing platforms often succeed in generating a list of possible medical conditions,
they fall short in providing actionable next steps. Specifically, they do not guide users to the
correct type of healthcare professional—such as a cardiologist, neurologist, or dermatologist—
based on their reported symptoms. This lack of direction can leave users confused, causing
delays in seeking appropriate treatment or consulting irrelevant specialists.

3. Barriers to Access and Affordability

Many AI-based health platforms operate on a freemium or subscription-based model, where
advanced features—such as detailed symptom analysis or personalized feedback—are hidden
behind paywalls. This business model restricts access to essential healthcare insights, particularly
for users from lower-income backgrounds or underserved communities. As a result, the digital
health divide continues to grow, further exacerbating healthcare inequalities.
4. Limited Localization and Contextual Relevance
A notable gap in most current solutions is the lack of region-specific customization. These tools
tend to deliver generic medical advice without considering geographic health trends, local
diseases, or regional healthcare infrastructure. For example, a chatbot trained on datasets from
Western countries may not accurately reflect the health challenges faced in developing regions.
This limits the relevance and reliability of their recommendations for global users.

5. Suboptimal User Experience and Usability

Ease of use is critical for the adoption of any digital health solution. However, many existing
platforms have interfaces that are cluttered, difficult to navigate, or designed with limited
consideration for users with low digital literacy. Furthermore, the lack of multilingual support
excludes non-English speakers, reducing the overall accessibility of the platform. These usability
issues create barriers that discourage regular use and limit the long-term impact of such tools.

Conclusion
These challenges underline the urgent need for a more intelligent, personalized,
and accessible healthcare assistant tool. A solution that combines machine learning
with symptom-specific guidance, offers specialist referrals, operates at no cost, and
reflects local healthcare contexts can bridge many of the gaps present in current
systems. The goal of this project is to address these limitations and deliver a more
inclusive and actionable AI healthcare experience.
 OBJECTIVES

The core objective of this project is to design and develop an AI-powered

healthcare assistant chatbot—DocTalk—that effectively addresses the major
limitations present in current digital health solutions. By combining machine
learning with intuitive design and equitable access principles, this project sets out
to deliver a functional and impactful tool that empowers users in managing their
healthcare needs. The specific objectives guiding this development are:

1. Develop a Reliable Diagnosis Engine

One of the foundational goals of this project is to build a robust and intelligent
symptom assessment system using a Decision Tree Classifier. This objective
focuses on:
 Training the model on diverse and relevant medical datasets to ensure high
accuracy.
 Designing the system to interpret user-reported symptoms and match them
to the most probable medical conditions.
 Enhancing the model’s adaptability to handle a variety of input
combinations for better diagnostic reliability.
The aim is to go beyond rule-based logic and leverage data-driven insights for
more personalized and context-aware diagnoses.

2. Integrate Specialist Recommendations

This objective is designed to bridge the gap between symptom identification and
actionable healthcare steps. Once a probable condition is determined, the system
will:
 Automatically suggest the most relevant type of healthcare professional
(e.g., cardiologist, dermatologist, neurologist).
 Educate users on the appropriate medical pathway based on their condition.
 Reduce uncertainty by offering structured, informed next steps instead of
vague or general suggestions.
This ensures users are not left guessing about what actions to take following the
chatbot interaction.
3. Design an Intuitive User Interface
To ensure accessibility for a wide range of users, including those with minimal
technical background, the chatbot will feature a user-friendly GUI built using
Tkinter. This objective involves:
 Creating a clean, responsive, and simple interface that mimics natural
conversation.
 Reducing cognitive load by avoiding clutter or overly complex design
elements.
 Supporting easy navigation and quick symptom input for seamless
interaction.
A strong focus on usability ensures the tool is approachable and inclusive.

4. Ensure Free and Equitable Access

Healthcare technology should not be a privilege limited to those who can afford it.
As such, one of the primary objectives is to:
 Make the chatbot completely free to use, without any premium barriers or
subscription requirements.
 Promote digital healthcare equity by enabling underserved and
economically challenged populations to access essential health guidance.
 Increase adoption across diverse user groups by removing financial entry
points.
This aligns with broader public health goals of accessibility and inclusiveness.

5. Enable Fast and Responsive Interaction

Timeliness is critical in healthcare support, especially for users seeking immediate
answers or guidance. This objective ensures that the chatbot is:
 Optimized for low-latency performance and real-time response generation.
 Capable of maintaining conversational flow without long delays.
 Designed to process user inputs efficiently and return results within seconds.
This responsiveness enhances the overall user experience and supports real-world
use cases where speed is crucial.

Summary
These objectives collectively shape DocTalk into a practical, user-focused
healthcare assistant. The project aims not only to enhance the technical
sophistication of symptom-checking tools but also to elevate the user experience
and widen accessibility, making reliable health support a reality for all.

TOOLS/PLATFORM USED
To ensure the successful development and smooth performance of the Health Care
Chat-Bot, a variety of programming languages, libraries, and development tools
were utilized:

Programming Language
 Python: Served as the core language for implementing backend logic, especially for
constructing and executing the decision tree classifier.

Frameworks and Libraries

 Tkinter: Used to design the graphical user interface, allowing users to interact with the
chatbot in a simple and intuitive manner.
 Scikit-learn: Employed to develop and train the decision tree model used for symptom
classification and prediction.
 NumPy & Pandas: Essential for handling, cleaning, and analyzing health-related
datasets to prepare them for model training.
 Matplotlib: Utilized to generate visual representations of data and model performance,
aiding in analysis and debugging.

Development & Deployment Tools

 Jupyter Notebook: Provided an interactive environment for testing, fine-tuning, and
visualizing model behavior during development.
 Visual Studio Code (VS Code): Served as the primary integrated development
environment (IDE) for writing and organizing project code.
 Git: Used for version control, collaboration, and efficient project tracking throughout the
development cycle.

External Resources
 Datasets: Symptom and disease data were stored in CSV format and used extensively for
both training and validating the machine learning model.

1. Overall System Architecture

 User Input (Symptoms)
↓
NLP Preprocessing (Tokenization, Lemmatization)
↓
Symptom Matching / Feature Extraction
↓
Decision Tree Classifier (Model Prediction)
↓
Diagnosis + Remedies + Specialist Recommendation
↓
Tkinter GUI Output to User

2. Machine Learning Workflow

 Dataset (CSV: Symptoms vs Diseases)
↓
Data Cleaning & Preprocessing (Pandas, NumPy)
↓
Feature Selection/Engineering
↓
Train/Test Split
↓
Model Training (Decision Tree - Scikit-learn)
↓
Accuracy Evaluation
↓
Saved Model for Prediction
 User Interaction Flow (via GUI)

User Launches GUI (Tkinter)

↓
Inputs Symptoms in Text Field
↓
NLP Engine Processes Input
↓
Model Predicts Disease
↓
Chatbot Suggests Remedies + Doctor Type
↓
User Acknowledges / Inputs New Query

Visual Image Representation of Tools

 [ Python ] —> [ Tkinter (GUI) ]
|
↓
[ NLP (Text Processing) ]
↓
[ ML Model (Decision Tree - Scikit-learn) ]
↓
[ Output: Recommendations via GUI ]

DESIGN METHODOLOGY
The development of DocTalk, an AI-powered health assistant, follows a structured
approach to ensure accuracy, usability, and responsiveness. The methodology
includes the following core components:

1. Symptom Analysis and Diagnosis

At the heart of the system lies the capability to accurately interpret and analyze user-reported
symptoms. This process includes:
 Symptom Input and Preprocessing
Users enter their symptoms through a simple, interactive interface built with Tkinter. The
input data is cleaned and normalized to remove inconsistencies, making it ready for
machine learning analysis.
 Prediction Using Decision Tree Classifier
Leveraging the Scikit-learn library, a decision tree model processes the preprocessed
symptom data to identify potential health conditions. The model is trained using labeled
datasets (Training.csv and Testing.csv) to ensure reliable and accurate predictions.

2. Specialist Recommendation System

Once a condition is predicted, the system guides the user by recommending the most relevant
medical specialist:
 Diagnosis-to-Specialist Mapping
A mapping mechanism links predicted illnesses with their corresponding specialties. For
example, respiratory symptoms point to pulmonologists, while dermatological concerns
are referred to dermatologists.

3. User Interface Design

The application features a user-friendly graphical interface aimed at accessibility and ease of use:
 Input Options
Includes a textbox for entering symptoms and dropdown menus for selecting additional
information such as related symptoms or user preferences.
 Result Display
The output section presents the predicted diagnosis along with the suggested type of
medical specialist.

4. Data Handling and Validation

Effective data integration ensures the model is both efficient and accurate:
 Feature Selection
Only the most significant symptoms are used for model training to enhance performance
and reduce noise.
 Validation Process
The dataset is split into training and validation sets to objectively assess the model’s
accuracy and prevent overfitting.

5. System Optimization
To maintain real-time responsiveness and efficiency, the system incorporates several
performance enhancements:
 Parallel Processing
Multiple user inputs are handled simultaneously using lightweight parallelization
techniques.
 Result Caching
Frequently accessed predictions are temporarily stored (cached), reducing processing
time for common queries and improving overall responsiveness.

USER INTERFACE
Dataset
Credentials
Code
 OUTCOME

As a result of the dedicated effort throughout our seventh-semester project, we

successfully developed DocTalk — a practical, efficient, and intelligent healthcare
chatbot designed to support preliminary health assessments. This AI-powered tool
serves as an accessible bridge between users and essential medical guidance,
especially in environments where professional healthcare may not be immediately
available.
The core achievements of DocTalk include:
1. Accurate Symptom Analysis
The chatbot is capable of analyzing user-reported symptoms with a high degree of
accuracy using a decision tree classifier, offering initial insights into potential health
conditions.

2. Specialist Recommendations
Based on the system’s diagnosis, users receive suggestions about the appropriate type of
medical specialist to consult, helping them take informed next steps in their healthcare
journey.

3. User-Friendly and Accessible Interface

With a clean and intuitive GUI built using Tkinter, the platform is easy to use for
individuals from both technical and non-technical backgrounds. Moreover, by offering its
services completely free of cost, DocTalk aims to eliminate financial barriers to basic
healthcare assistance.

Beyond these features, DocTalk demonstrates the potential of AI in democratizing

healthcare. It empowers users by providing fast, personalized, and actionable
advice—anytime, anywhere. While not a replacement for professional medical
consultation, it serves as an effective first step for individuals seeking preliminary
guidance, particularly in underserved areas or during early symptom onset.

This project reflects our commitment to leveraging technology for social good and
highlights how machine learning and thoughtful design can come together to
address real-world challenges in healthcare access.
 GANTT CHART

TASKS WEEK WEEK WEEK WEEK 4 WEEK WEEK WEEK 7

1 2 3 5 6

defining
problem
stateme
nt

assignin
g tasks

Data
collectio
n and
preproc
essing

Building
decision
tree
model

designi
ng GUI

Model
integrati
on with
GUI

Testing
and
validatio
n
 RESPONSIBILITY CHART

TASK Harshit(tea Mentor

m
member)

defining R I
problem
statement

assigning R I
tasks

Data collection R I
and
preprocessing

Building R I
decision tree
model

designing GUI R I
and
Integrating
model with
GUI

Testing and R I
validation
 REFERENCES

1. Scikit-learn Documentation: https://scikit-learn.org/

2. Pandas Documentation: https://pandas.pydata.org/
3. NumPy Documentation: https://numpy.org/
4. Tkinter Official Guide:
https://docs.python.org/3/library/tkinter.html
5. Matplotlib Guide: https://matplotlib.org/
6. Python Official D
7. ocumentation: https://docs.python.org/3/
8. Training and Testing Dataset: Provided as CSV files
(Training.csv and Testing.csv).
 Annexure I:

Front page of plagiarism report by guide