MAJOR-1 PROJECT

Documentation
For

Chatbot Based on AI
Submitted By

Specialization SAP ID Name

AIML 500096249 Aryan Mankotia

AIML 500096806 Shubham Mehta
AIML 500096748 Priyanshu
Gupta

AI Cluster
School of Computer Science
University of Petroleum & Energy Studies
Dehradun - 248007. Uttarakhand
Project Guide Cluster Head
Index

S. No. Title Page No.

Abstract 3
1 Introduction 3
2 Literature Review 4
3 Problem Statement 4
4 Objective 5
5 Methodology 6
6 Results 8
7 Future Scope 9
8 Conclusion 9

2
 Synopsis Report

Project Title
Chatbot Based On AD

Abstract
This project develops an advanced chatbot system for Alzheimer's disease prediction, using
convolutional neural networks (CNN) to analyze brain MRI images. The chatbot incorporates
image-based predictions for early detection, as well as natural language processing (NLP) to
engage with users through text-based interaction. Additional features include Wikipedia
integration for contextual information and sound-based interaction via speech recognition,
offering cognitive assessments through auditory inputs. The user interface is designed to be
intuitive, providing an accessible platform for uploading images, conversing with the chatbot,
and receiving diagnostic feedback. This multi-modal system combines image, text, and sound
for a comprehensive Alzheimer’s prediction and support tool.

1. Introduction
Chatbots have become indispensable tools in recent times for improving user interaction in a
variety of industries, such as customer service, education, and healthcare. Intelligent
conversational bots are in high demand as companies look for ways to increase productivity
and user happiness. Users become frustrated with traditional rule-based chatbots because they
frequently are unable to provide meaningful and context-aware interactions.
The Alzheimer's Disease Support Chatbot is designed to assist caregivers and patients in
managing the challenges associated with Alzheimer's. This AI-driven system provides timely
information, emotional support, and personalized care guidance, enhancing the quality of life
for both patients and caregivers. By using natural language processing (NLP) and machine
learning, the chatbot can offer practical advice on daily activities, symptoms, and coping
strategies while fostering engagement and reducing feelings of isolation. It aims to be a
reliable companion, promoting patient safety and caregiver well-being through accessible,
compassionate, and continuous support.

Convolutional Neural Networks (CNNs) are used in Alzheimer's disease research for
analyzing brain imaging data, such as MRI and PET scans. By automatically detecting
patterns and abnormalities, CNNs help in early diagnosis and progression prediction,
improving the accuracy of identifying Alzheimer's-related brain changes and aiding in timely
intervention and treatment.

The Wikipedia Library enhances chatbot systems by integrating its extensive and reliable
knowledge base, enabling chatbots to provide accurate and detailed responses. This
integration improves the chatbot's ability to handle a wide range of inquiries with well-
informed answers. For sound systems, text-to-speech technology can be used to vocalize the
content, making information accessible audibly. The interface design benefits from seamless
integration, allowing users to query directly through the chatbot and receive structured,
relevant responses. This combination enriches the user experience, making interactions more
informative and engaging.

2. Literature Review
3
The authors in [1] proposed an innovative AI-driven solution: employing an LLM based
chatbot as an alternative to human interviewers in dementia intervention programs. LLMs or
Large Language Models are state-of the-art Machine Learning models that can conduct open
conversations like humans. Compared to human counterparts, LLM-based chatbots are
accessible round-the-clock at lower costs. The chatbot was specifically designed to offer
cognitively demanding conversations, mimicking the human-interviewer-based previous, and
providing patient-friendly interfaces to effectively engage with older adults Furthermore, it
ensured natural, effortless conversations through a soothing vocal interface, thereby
accommodating older adults with varying cognitive capacities.

The authors in [2] implemented and trained 4 models including Long Short-Term Memory
(LSTM), Gated Recurrent Unit (GRU), Bidirectional LSTM, and Dense Neural Network on
preprocessed dataset of first aid. Performance evaluation metrics such as accuracy, precision,
recall, and user feedback were employed to assess the efficacy of the developed chatbot. The
development process involved several key phases. Initially, a dataset comprising first aid
intents is collected and meticulously preprocessed using NLTK's tokenization, part-of-speech
tagging, and syntactic parsing functionalities to extract meaningful information from user
inputs. The DNN was best performing.

In [3] the authors have distinguished between two classes: mild-demented and non-demented.
CNN was used because its able to handle enormous datasets and achieve good classification
performance. The authors proposed two architectures for the study: A pretrained VGG 16 and
a CNN model built from scratch. The data was augmented by six operations: rotation at 15
degrees, width shift with a range of 0.1, height shift with a range of 0.1, shearing with a range
of 0.2, zooming with a range of 0.2, and applying horizontal flip CNN gave better accuracy
than VGG 16.

In [4] the authors provide a comprehensive survey on the evolution and deployment of LLM-
based AI chatbots, particularly highlighting their impact since the advent of OpenAI's
ChatGPT. It covered the development of foundational chatbots, the progression of large
language models (LLMs), and their applications across various sectors including education,
healthcare, software engineering, and finance. The study emphasizes the potential of these
chatbots in providing medical guidance, software development support, and other industry-
specific solutions, showcasing their efficiency and scalability.
In [5] the authors proposed Artificial Neural Networks (ANNs) are powerful for capturing
complex, non-linear patterns using ReLU activation in hidden layers and Softmax for multi-
class classification. Logistic regression was simpler but assumes linearity, making it less
effective for complex data. Decision trees handle non-linearity but tend to overfit, while
random forests reduce overfitting by aggregating multiple trees. Support Vector Machines
(SVMs) excel in high-dimensional spaces but are computationally intensive. Each model
offers different strengths based on task complexity and computational needs.
In [6] the authors provided an approach which divided chatbot development into information
abstraction and response design. Information abstraction involves gathering data, applying
Natural Language Processing (NLP), and using machine learning algorithms like
classification techniques to structure and organize queries. Data manipulation stores and
optimizes this information for future interactions, using tags and database management
systems like MySQL or MongoDB. Response design leverages tools like Dialogflow to
4
configure intents, contexts, and entities, enhancing dialogue flow and user experience.
Finally, entity recognition and custom objectives improve chatbot accuracy by better
understanding and grouping user interactions.
In [7] The authors discussed the development of a novel medical chatbot designed to
diagnose Alzheimer's disease using a Deep Neural Network (DNN). The chatbot allowed
users to interact with an AI system to analyze symptoms and MRI scan reports, providing a
more accessible and timely diagnosis compared to traditional methods. The system aimed to
reduce the need for frequent hospital visits, especially for working individuals. The chatbot
leverages machine learning techniques, including TensorFlow and Keras, to process data and
improve prediction accuracy. The research highlighted the chatbot's potential to function as a
virtual doctor, offering diagnostic support and analyzing disease severity, symptoms, risk
factors, and prevention methods.
The authors in [8] implemented a structured prototype methodology, the chatbot was
designed to improve information retrieval and provide personalized article recommendations.
User Acceptance Testing (UAT) yielded a high satisfaction score of 4.14, indicating positive
user experiences. The chatbot's integration with the OpenAI API significantly improved
information extraction and user satisfaction.
The authors in [9] implemented ChatterBot library and hosted on the web using the Flask
framework. The chatbot begins by collecting the user's name, primary ailment, and duration
of symptoms, followed by more detailed symptom inquiries. The responses are processed and
stored in a database, and the chatbot provides answers based on the highest accurate match
from the database. The paper emphasizes the chatbot's ability to offer quick and accurate
responses, making healthcare more accessible, especially for those who cannot afford
frequent doctor visits.

In [10] the authors outlined the evolution of digital healthcare, from the digitization of patient
records to the use of mobile apps and wearable devices for health monitoring. It also
emphasizes the potential benefits of LLMs in improving medical processes, providing
evidence-based treatment options, and enhancing patient access to medical information.

In [11] the authors discussed the importance of thoroughly testing, verifying, and validating
chatbots to prevent failures, especially when faced with ambiguous or illegible inputs. It
emphasized on algorithm testing as a promising solution, exploring techniques like cross-
validation, grammar and parsing verification, and statistical parsing. The paper details the use
of Naive Bayes and Support Vector Machines (SVMs) for text classification and intent
identification, highlighting their testing methods and performance metrics. The goal was to
enhance chatbot reliability and make interactions indistinguishable from human
conversations.
In [12] the difference between the brain of patients and brain of normal persons is shown. In
AD the shrinkage of cerebral cortex happens along with shrinkage of hippocampus and
enlarged ventricles in between.
In [13] the authors used the Cornell Movie Dialog Corpus, containing over 220,000
conversation exchanges from 617 films to develop a chatbot based on the BiLSTM model.
This chatbot processed user inputs to generate movie-related dialogue responses by
leveraging bidirectional LSTM’s ability to capture both past and future context. The chatbot's
5
architecture included sequence-to-sequence translation methods, and its output is evaluated
based on the accuracy of its responses. Embeddings and feature-based models helped manage
sparse data by extracting word characteristics, improving the chatbot's performance.
Evaluation ensured the chatbot’s generated responses align with user inputs effectively.

3. Problem Statement
Alzheimer's disease presents major challenges for loved ones as well as patients because of
its progressive nature and extensive management requirements. Current Alzheimer's patient
support systems do not provide real-time, interactive, or personalized assistance. This gap can
result in insufficient symptom management, higher stress levels for caregivers and a lower
quality of life. A chatbot created expressly for Alzheimer's disease could help bridge this gap
by providing timely information, emotional support, and practical advice. The lack of such a
specialized tool emphasizes the need for novel solutions to improve patient and caregiver
experiences when managing Alzheimer's disease.
4. Objectives
The objective of this project is to develop a chatbot tailored to Alzheimer’s disease that
provides personalized support and information to patients and caregivers. The chatbot will
aim to enhance patient well-being by delivering real-time assistance, tracking symptoms,
offering cognitive exercises, and providing resources for managing daily activities.
Additionally, it will support caregivers by offering stress-relief strategies, caregiving tips, and
connecting them with relevant support networks. By integrating AI and natural language
processing, the chatbot will improve the quality of care and support for individuals affected
by Alzheimer’s disease

5. Methodology
5.1. Image Data Collection:
o Obtain Alzheimer's disease image datasets from reliable sources, ensuring a mix
of modalities (MRI, CT scans).
o Normalize and resize images, then split into training, validation, and test sets.
5.2. Model Development:
o Model Selection: Choose best CNN architectures
These steps have been completed for 4 stages of Alzheimer's disease across the mentioned
CNN models.
5.3. Chatbot Development
5.3.1. NLP based Chatbot Creation

a) Requirement Analysis:
 Define the scope and functionalities of the chatbot (e.g., symptom checker,
general information on Alzheimer’s).
b) Dataset for Chatbot Training:
 Collect and curate a dataset of common questions and responses related to
Alzheimer’s disease.
 Include varied user intents and contextual responses.
c) Text-based Response Generation:
 Preprocess user query, vectorize using TF-IDF, and compute cosine similarity
to select the most relevant response.
6
  Display the relevance scores too.
d) User Interaction Design:
 Design and implement the chatbot interface for user interaction using PyQt5.
 Ensure the interface is intuitive and user-friendly.

5.3.2. Enhancing Chatbot Features

a) Sound Features:
 Integrate text-to-speech (TTS) functionality for voice responses.
 Add sound effects to enhance user engagement and experience.
b) Visual Interface:
 Develop a graphical user interface (GUI) with options for text and image inputs.
 Include visual elements like buttons, images, and interactive components for a
richer user experience.
 Image Analysis: Preprocess the MRI image, extract features using a CNN,
generate a caption describing the brain scan, and predict the stage of Alzheimer's.
5.4. Testing and Validation: -
a) System Testing:
 Conduct comprehensive testing to ensure the chatbot functions correctly and
integrates seamlessly with the CNN model.
 Perform usability testing to identify and address user experience issues.
b) Feedback Collection:
 Gather feedback from users to identify areas for improvement.
 Iteratively refine the chatbot based on user feedback and performance metrics.
5.5 Deployment and Maintenance: -
a) Deployment:
 Deploy the chatbot on relevant platforms (e.g., web, mobile applications).
 Ensure it is accessible to target users and performs reliably in real-world
scenarios.
b) Maintenance:
 Regularly update the chatbot to accommodate new information and
improvements.
 Monitor performance and address any issues or bugs that arise.

6) Results:

7
Fig 6.1 Shows output/ caption for MRI image analysis

Fig 6.2 Shows The Home Screen Of Interface

8
Fig 6.3 Shows answer to a textual query entered by the user.
7) Future Scope
The future scope of this project is broad, and includes improving the system’s accuracy,
expanding and updating dataset regularly making it more interactive and accessible,
integrating it into healthcare workflows, enhancing user engagement, and expanding its reach
to broader use cases, including monitoring other neurological diseases. These improvements
would allow the system to have a larger impact, helping both patients and healthcare
providers in the early detection, diagnosis, and management of Alzheimer’s disease and
related conditions.
8) Conclusion
The integration of advanced deep learning models for image analysis, natural language
understanding using more sophisticated NLP models, and additional features like voice
interaction and multilingual support enhanced the chatbot's capabilities. Moreover,
integrating the system into real-world healthcare workflows through EHR integration and
clinical validation will be crucial for its wider adoption in clinical settings.
In conclusion, this project serves as a foundation for a powerful tool that could improve the
early detection, management, and overall care of individuals affected by Alzheimer's disease.
It bridges the gap between artificial intelligence and healthcare, offering a scalable solution
that could transform the way Alzheimer's is understood and managed in both clinical and
home environments. With ongoing improvements and further research, this system could
become a key player in the global effort to fight Alzheimer’s and similar neurodegenerative
diseases.

References
[1] Hong, J., Zheng, W., Meng, H., Liang, S., Chen, A., Dodge, H. H., ... & Wang, Z. A-
CONECT: Designing AI-based Conversational Chatbot for Early Dementia Intervention. In
ICLR 2024 Workshop on Large Language Model (LLM) Agents.

9
[2] Kumar, T. R., Akarapu, A., Allam, S. R., Gattepally, N. C., & yashaswi Nellutla, P.
(2024). CHATBOT APPLICATION USING NLTK AND KERAS.
[3] Arafa, D. A., Moustafa, H. E. D., Ali, H. A., Ali-Eldin, A. M., & Saraya, S. F. (2024). A
deep learning framework for early diagnosis of Alzheimer’s disease on MRI images.
Multimedia Tools and Applications, 83(2), 3767-3799.
[4] Dam, S. K., Hong, C. S., Qiao, Y., & Zhang, C. (2024). A complete survey on llm-based ai
chatbots. arXiv preprint arXiv:2406.16937.
[5] M. Jain, P. Nathe, K. Rathod, N. K. Tiwari, S. Dedgaonkar and C. Shewale, "AI
HealthCare Chatbot," 2024 MIT Art, Design and Technology School of Computing
International Conference (MITADTSoCiCon), Pune, India, 2024, pp. 1-6, doi:
10.1109/MITADTSoCiCon60330.2024.10575622. keywords: Ethics;Data privacy;Medical
services;Artificial neural networks;Chatbots;Data models;Security;Chatbot;Machine
Learning;Deep Learning;Artificial Intelligence;HealthCare;Artificial Neural Networks
[6] P. Tanwar, K. Bansal, A. Sharma and N. Dagar, "AI Based Chatbot for Healthcare using
Machine Learning," 2023 International Conference on Advances in Computation,
Communication and Information Technology (ICAICCIT), Faridabad, India, 2023, pp. 751-
755, doi: 10.1109/ICAICCIT60255.2023.10465728. keywords:
{Industries;Automation;Natural languages;Medical services;Machine learning;Oral
communication;Chatbots;Machine Learning (ML);Artificial Intelligence
(AI);Chatbot;Natural Language Processing (NLP) Dialogflow},
[7] P, M., V, A.D., D, B., & S, M. (2023). A Novel Medical Chatbot with Alzheimer’s Disease
Detection Using Deep Neural Network. Computational Intelligence and Machine Learning.
[8] Santoso, G., Setiawan, J., & Sulaiman, A. (2023). Development of OpenAI API Based
Chatbot to Improve User Interaction on the JBMS Website. G-Tech: Jurnal Teknologi
Terapan, 7(4), 1606-1615.
[9] Goel, A., & Sharma, S. (2023, June). Artificial Intelligence based Healthcare Chat Bot
System. In 2023 8th International Conference on Communication and Electronics Systems
(ICCES) (pp. 1362-1366). IEEE.
[10] Park, S. G., Kim, A., Yoon, T., Kamyod, C., & Kim, C. G. (2023, November). A Study
of Generative Large Language Model for Healthcare. In 2023 7th International Conference
on Information Technology (InCIT) (pp. 397-400). IEEE.
[11] Vijayaraghavan, V., & Cooper, J. B. (2020). Algorithm inspection for chatbot
performance evaluation. Procedia Computer Science, 171, 2267-2274.
[12] Breijyeh, Z., & Karaman, R. (2020). Comprehensive review on Alzheimer’s disease:
causes and treatment. Molecules, 25(24), 5789.
[13] P. Anki, A. Bustamam, H. S. Al-Ash and D. Sarwinda, "High Accuracy Conversational
AI Chatbot Using Deep Recurrent Neural Networks Based on BiLSTM Model," 2020 3rd
International Conference on Information and Communications Technology (ICOIACT),
Yogyakarta, Indonesia, 2020, pp. 382-387, doi: 10.1109/ICOIACT50329.2020.9332074.
keywords: {Recurrent neural networks;Impedance matching;Chatbot;Data processing;Data
models;Information and communication
technology;Python;chatbot;program;BiLSTM;accuracy;input;output},

10