AI-Driven Chatbot for Mental Health Support Using Sentiment

Analysis

V. Mythily, D. Vinoparkavi, P. Sukumar, Chanchalhas V, Ajith M and Barani Kumar K M

Department of Computer Science and Engineering, Nandha Engineering College, Erode, Tamil Nadu, India

Keywords: Mental Health, Sentiment, Analysis, Health Support.

Abstract: The Mental care remains a problem due to its low accessibility. This paper proposes AI-controlled chatbots

based on mood analysis to identify emotions and submit appropriate answers. The system uses NLP and

machine learning to provide emotional care in real-time and self-help materials. The system is confidential,

scalable and provides early intervention for mental wells. For the majority of the population, psychological

health care is still not easy for the surrounding stigma. The proposed paper provides an AI-driven chatbot that

recognizes emotions and provides appropriate responses through mood analysis. The system integrates natural

language processing and algorithms for machine learning to provide real-time emotional support for self-

therapy and provide resources for self-therapy. Created for privacy scalability and affordability the bot offers

a subtle, private midfield for anyone who wants to receive emotional support. Then it may be the first sign of

already needing it and serves as an early intervention tool that may be useful if it is needed. The fusion of

gaps between users and healthcare is an easy first step for those who are unsure whether they are willing to rely

on what is ultimately perceived as professional help. The experimental results show that the AI-operated

chatbot developed here can perform mood analysis with satisfactory quality and recognize emotional nuances

while having empathetic conversations. The system generates corresponding answers that are intended to

provide comfort and support, not just capture the detection of emotional signs from user statements. It uses

techniques such as relaxation exercises and CBT to focus on aggressive reinforcement. It is based on general

principles of mental health problem solving with the aim of strengthening concerns for users. The ability to

adapt is important to enable timely and real-time responses based on the user's emotional state. Using adaptive

learning characteristics this tool based on emotional variables, allows flexible adaptation to a single user in

real time. User interactions develop into useful ones that can contribute to treating emotional problems under

untreated conditions. Chatbots cannot replace therapy or specialized care. It is easy to achieve the initial

intervention for those looking for support before full treatment with a clinician. Chatbots adapt to user

interaction in real time to improve understanding of emotional information and provide an appropriate support.

It also serves as the first treatment tool for mental health to help people before specialized treatment. Future

work will include developing mood analysis models and supporting multilingual support for more people.

1 INTRODUCTION

This Mental well-being is an important aspect of the

well, but the majority of the population is

inaccessible in time due to stigmatization, lack of

resources and financial constraints. With advances in

technology, especially artificial intelligence, AI-

based chatbots have now become a practical solution

to provide scalable and accessible intellectual well-

being. This article presents an AI-based chatbot. This

uses mood analysis to implement the user's emotional

well-being and responds accordingly. Chatbots can

identify emergency stress patterns and provide

personalized support when processing natural

language processing (NLP) and machine learning.

Provide real-time talk, self-help materials and move

to expert support if necessary. Chatbots are

confidential, available at any time, and provide

mental well-being. This system is very effective for

early intervention and emotional wells. Experimental

results show that they are effective in detecting

emotions and sensitive feedback. The developed

chatbot aims to bridge the gap between people and

mental health care with AI-based technology. The

292

Mythily, V., Vinoparkavi, D., Sukumar, P., V., C., M., A. and M., B. K. K.

AI-Driven Chatbot for Mental Health Support Using Sentiment Analysis.

DOI: 10.5220/0013881700004919

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 1st International Conference on Research and Development in Information, Communication, and Computing Technologies (ICRDICCT‘25 2025) - Volume 2, pages

292-297

ISBN: 978-989-758-777-1

application of AI for mental health interventions has

the ability to change the delivery of emotional

support from data-controlled real-time feedback.

This study examines the possibilities of mood analysis

in improving the mental well-being supported by

chatbots.

Artificial intelligence (AI) transforms mental

health care through real-time, data-driven emotional

feedback. Chatbot interventions, facilitated by mood

analysis-based interventions, offer the potential for

targeted mental health interventions with

thepossibility of recognizing emotions, pursuing

mood patterns and intervention appropriately. This

study applies AI-based mood analysis to examine the

potential for rapid, accessible and effective mental

health mood changes to chatbot communications in

responses. Such a real-time feedback loop maximizes

commitment, reduces loneliness, and allows for early

intervention. AI-driven chatbots for mental health are

economical and scalable solutions to bridge the gap

between traditional treatments and traditional

treatments everywhere. This study will examine the

feasibility, validity and ethics of mood analysis in

chatbot-based mental health interventions. The

results may inform the design of AI models that

provide sensitive and human support with user

privacy and ethics certainty. The aim of this study is

to investigate the feasibility, reliability and moral

impact of injecting mood analysis in intellectual care

via chatbots. This finding could lead to the creation

of AI models that provide caring, natural support that

maintains user privacy and at the same time respects

ethical limitations.

2 RELATED WORKS

Many AI chatbots have been adopted, as they are

perceived to play an instrumental role in mental

health care: they assess users' feelings and offer

treatment responses individualized to that mood.

They operate by the principles of natural language

processing and machine learning to identify the

signals of emotional distress, with the intention and

ability to respond in an empathic manner. Many

studies have evaluated the effectiveness of AI

chatbots in terms of quickness of interventions or

user engagement. Being able to make correct

sentiment-assessment and eliminate bias in the AI

model are key success factors influencing their

effectiveness.

Incredibly AI chatbots are now steadily being

utilized in mental healthcare and highly leverage

sentiment analysis and AI to provide assistance. Their

research suggests real-time intervention as well as

easy accessibility. These supporting studies agree

with the belief that chatbots can deliver cognitive-

behavioural therapy more efficiently and reduce

anxiety symptoms through mobile applications.

Systematic reviews agree that conversational agents

would enhance patient engagement but struggle with

matters of accuracy and ethics Fadhil, A., & Moffatt,

K. (2020).

Fitzpatrick et., al. (2017). AI chatbots

contribute to the psychological well-being of people

by providing personalized mental health support.

Nevertheless, reliability and data privacy issues are

paramount towards the success of chatbots. Future

directions will be to further enhance intelligence and

ethical deployment of the chatbot.

AI-enabled chatbots have changed mental

healthcare mostly by making it more available and

also by providing timely assistance. The literature

suggests they are effective in the delivery of

cognitive behavioural therapy through mobile

applications that reduce symptoms of anxiety and

depression. Studies show that conversational agents

contribute to better patient engagement as well as

facing challenges such as ethical challenges data

privacy and accuracy concerns (Laranjo, L., et al.

(2018)).

(Tielman, M., & de Vries, L. (2020)).

These AI-Driven chatbots play a role in

psychological well-being by allowing personalized

mental health support. However, the promise of

combining AI and mental health care also has issues

that need to be dealt with for a successful

implementation. The future developmental agenda

would be on making the bots smarter and ethically

deployable in mental health contexts.

The chatbots powered by artificial intelligence

play an increasingly important role in mental

healthcare by providing accessible, real-time support

for individuals who are suffering from anxiety and

depression. Research shows that they work

exceptionally well in delivering mental health

interventions, aiding patient engagement, and

providing self-help tools.

(Miller, C. J., et al. (2020)).

These chatbots use AI-driven sentiment analysis to

diagnose the user's emotional state and provide

personalized responses, which can ultimately

enhance the therapeutic experience. However, ethical

considerations, security of data, and accuracy remain

very important issues. By overcoming these barriers

chatbot trustworthiness could further improve and

make for a better user experience in mental health

applications. Such advancements should focus on

improving AI models to enhance emotional

intelligence toward the ethical generation of mental

health care services

(Lee, J. A., & Choi, J. (2020)).

AI-Driven Chatbot for Mental Health Support Using Sentiment Analysis

293

AI-powered chatbots have developed rapidly into

very efficient digital tools within the sphere of mental

healthcare. In this way, they offer in-the-moment

support and intervention to those suffering from

psychological distress

(Lau, A., & Wenzel, S.

(2019)

). (Radzi, S., & Khamis, M. (2020)).

Literature recognizes a role for chatbots in promoting

patient engagement through personalized

interactions and digital interventions tailored to

individual needs. (Gaffney, H., & Kuss, D. (2021))

Studies suggest that they could be put to good use in

providing therapy and self-help strategies aimed at

better mental health outcomes. (Ly, K. H., &

Andersson, G. (2020)) Data privacy issues, the ethical

implications, and the accuracy of AI-based responses

remain some of the critical challenges. Weitzman, (E.

R., & Parikh, R. (2019)) Reliability and more humane

AI interaction can help to promote trust and grow

intervention credibility through chatbot-based mental

health approaches. (Li, Y., & Wang, T. (2020)) Most

of all, in the future, the focus should be on improving

the AI algorithms for better and deeper emotional

understanding and ethical deployment within mental

healthcare.

3 PROPOSED METHODOLOGY

& IMPLEMENTATION

Figure 1: Chatbot architecture diagram.

The planned AI-powered chatbot solution is based on

multi- phase development, which should

appropriately support mental well-being via

sentiment analysis (figure 1). First, the chatbot gives

the users the opportunity to speak their thoughts,

feelings, or issues while maintaining a safe yet

supportive environment. The textual input is then

subjected to various methods of pre-processing

during cleaning, such as tokenization, removal of

stop words, stemming, lemmatization, and cleaning

up of special characters in order to be ready for

analysis.

The sentiment analysis model uses machine

learning or deep learning method BERT to assign one

of the three sentiments positive, negative or neutral.

The identified sentiment and emotions help deliver

the pre-defined templates or provide the generative

AI model with the responses in a good and warm

conversational manner. These models can be fine-

tuned with mental health specific data sets to pick up

subtle emotions pertaining to the mental world. This

allows the sentiment analysis model to classify

emotion in the conversations, thus allowing the

chatbot to figure out the user’s mental state.

Emotional overload topics, such as answering tone

and language will have frameworks of agreed-upon

responses so that the system will recognize emotions

in the user. Such tasks as manifesting the nuances of

users' inquiries including the identification of

particular mental health issues or triggers mentioned

by the user will be undertaken partly through natural

language processing methods (NLP). Another area

for development is dialog management which

enables a chatbot to follow a coherent context aware

conversation ensuring that its response fits into the

flow of dialogue and appears appropriate for the

emotional experience of the user.

3.1 Data Collection and Preprocessing

An AI-based mental health chatbot is basically

designed on the basis of structurally arranged data

pulled out from conversations chat logs and surveys

distributed among users. In order to enhance the

accuracy of the analysis of the sentiments from the

social media posts text pre-processing and cleaning

has had to be done on the processed text (table 1).

Text cleaning eliminates unwanted characters,

symbols and numbers.

Lowercasing makes it uniform while tokenization

separates it into meaningful units. Stop word removal

eliminates irrelevant words while lemmatization

changes the words into their base form. Beyond this

further refinement extracted the punctuation did spell

correction and expanded slang. Dealing with

negation would carry out a phrase such as "not

happy" to be interpreted as negative and better

User Input

User Interface

Preprocessing

Classification

Sentiment

Ali

User feedback to

rove

Suggestion

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

294

chatbot responses especially empathy when needed

all for mental health support.

Table 1: Data Collection and Preprocessing Table.

STEPS DESCRIPTION EXAMPLE

Data Collection Gather text data

related to mental

health

I'm feeling

really down

today and it's

just awful!

Data Cleaning Remove irrelevant

text, and special

characters, HTML

tags and stop words to

make the data useful.

Converting I'm

feeling very sad

depressed to

I'm feeling very

sad

Tokenization Break text into an

individual word.

I feel anxious

today → I,

feel, anxious,

toda

3.2 Data Integration

Data integration is paramount for the AI mental

health chatbots, as combining text, speech and

physiological data presents the opportunity for more

personalized responses. The application of an array of

data sources will expand the user's emotional insight,

thereby enabling accurate support. Different data

type’s integration helps to improve the chatbot's

ability to perceive subtle emotional shifts and results

in an interaction of high calibre with the user

empathizing responses lead to a more enhanced user

experience. Here the issues of privacy concern,

interface compatibility and real-time processing

requirements awaiting decision must be addressed for

seamless and secure data transfers and the future

work will focus on building the efficacy of the

methods of integration as well as extending the

chatbot's ability to process varied data inputs on

mental health care to gain more effective supportive

action.

3.3 Emotion Detection

We did it with a Pre-trained BERT Model Fine-

Tuning on Emotion Classification supported by Go

Emotions dataset in which emotional responses are

categorized into 27. It classifies user input into one of

those categories happy, sad, and stressed, etc. The

emotion detection model processes the input text and

upon detecting the emotion and the chatbot picks the

appropriate response.

3.4 Motivational Chat bots

Motivational chatbots have been designed as entities

that provide encouragement to elevate mood and

others as mental support through reaffirms or

relaxation. Breathing exercises, self-reflexive-

questioning or positive affirmations are typically

offered as greater solutions based on the recognized

emotional state.

The demand for instant emotional support

chatbots is on the rise these AI Mental Health

Assistants passively absorb warmth and empathy

from human interaction-oversaturation an attempt to

mimic that ineffable quality of direct connection.

Reactions range from pre-programmed lines to

friendly hope inspired suggestions. While quite

invaluable though that they do not usually take into

consideration the more extreme nuances of emotional

issues. Basically, AI still hasn't perfectly launched the

delivery of full individualism while working with

chatbots some might sometimes miss out on

understanding subtle emotions. AI- enabled chatbots

are settling into their place as adjuncts to traditional

therapy for mental health. They have made it possible

to give a non-threatening space to clients who feel

frightful about seeking professional help thus

granting them support and a degree of emotional

modulation. By bridging the gap towards immediate

emotional relief in conjunction with mental health

long-term wellness strategies through primary

professional care, the chatbots might continue to

minimize the chasm. Essentially, AI-based chatbots

cannot substitute human connect their role in support

of nurtured emotional support is maturing, slowly

making mental health resources more accessible more

engaging and responsive to individual needs.

Through 24 hour a day support these chatbots help

people deal with different emotional challenges

practically in real time. They can’t no replace

professional caregivers but they boost self-awareness

and emotional resilience in a good way.

3.5 Chatbot Workflow

The mainstream steps in the workflow of chatbots are

layered starting with the user interaction layer where

a user sends some message to the chatbot indicating

a thought or feeling. The chatbot ingests this input

data into its logs saving it for subsequent

processing. More specifically this should engage in

tokenization stop word removal and lemmatization.

Here the structure of sanitization allows any further

work of the chatbot. Then the Emotion and

Sentiment Detection Layer uses a sentiment

AI-Driven Chatbot for Mental Health Support Using Sentiment Analysis

295

analysis model likely BERT to predict emotion from

the input data. Such states as positive (happy,

motivated), neutral (calm, reflective) or negative

(anxious, stressed or depressed) are detected. With

the emotional state identified, the internal

mechanism goes to the Response Generation Layer,

where it chooses an appropriate response to give. If

One's frozen type of interaction gives a positive signal

then the chatbot responds with positivity and a

rooting approach. If from one's frozen interaction

type, engagement is neutral then a normal chatter

ensues. In the worst-case situation the chatbot

dispenses some motivation utility, relaxation

exercise or coping skills. The User Emotional Trend

Tracking Layer tracks user activities to identify

trends of emotional patterns over time.

Table 2: Model Evaluation and Optimization.

Evaluation

Metric

Description

Optimization

Strateg

Accuracy

Measures how

often the model

correctly

classifies user

sentiment.

Fine-tune BERT

using domain-

specific datasets

Precision

Percentage of the

positive

predictions that

are correct.

Use class balancing

to avoid false

positives.

Recall

(Sensitivity)

To Measures the

ability to detect

the emotional

distress more

correctly.

Adjust threshold

tuning for better

sensitivity.

F1-Score

Balances precision

and recall for the

better and

overall

erformance.

Use cross-

validation to fine-

tune

hyperparameters.

This seems cyclic to determine if there is a

consistent report of pain persisting to warrant another

change by the chatbot. In Response Delivery Layer

the response is directed to the user with an

engagingly conscious emotional awareness within

the conversation. Thus arises the escalation call for

referral resources of mental health professionals and

emergency support contacts in the event of an

escalating serious issue. It's very apparent that the

assistance by an adaptive real time sense- making

based mental health chatbot exists for immediate

support. Table 2 shows the model evaluation and

optimization. Figure 2 depicts the emotion detection

accuracy (%) and figure 3 depicts the performance of

emotion prediction model.

Figure 2: Emotion Detection Accuracy (%).

Figure 3: Performance of emotion prediction model.

4 RESULT

An AI-driven chatbot was built to analyse user

emotion through the technique of sentiment analysis.

The chatbot successfully identified the positive,

neutral and negative emotion with an approximate

accuracy of 85% while the chatbot was able to

effectively indicate instances of a distress signal from

the users and prompted a follow-up for emotional

support from the users the chatbot needs to work on

its efficiency for scenarios that are indeed complex.

Future development must work on the issues of

sensitivity to emotions, bias mitigation, and ethical

development of any AI managing sensitive

environments in respect of mental health support

needs.

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

296

5 CONCLUSION AND FUTURE

WORK

The clinical strength of this AI-assisted chatbot for

mental health support is innovative and accessible by

humans experiencing stress, anxiety and emotional

distress. These chatbot helps us by keeping our

mental health in normal level and helps during

critical and hard situations like stress, mental

pressure. Capable of interpreting feelings from a text

form through advanced neural networking such as

BERT-based models this chatbot accurately

determines emotion-detection logics and packs

motivational messages and keeps our mental health

normally in an healthy way and visualization relaxation

techniques as solutions. It is distinctly different from

the previous chatbots in that the present-day chatbot

can store the changes of emotion through time,

thereby paving for some meaningful interactions and

adaptation. It is useful for a quick response in

eventualities relative to mental health care issues. The

chatbot will encourage self-reflection and emotional

awareness, thus instilling consciousness for

longevity. It is built to scale allowing multi-language

support and expansion to a higher reach. Future

improvements will be geared toward improving

contextual understanding and learning. This project

shows how AI can contribute to the improvement of

any mental health issues and empathetic

interventions in collaboration with human views.

Future work includes working on improving the

multi-label emotion detection for better recognition

of overlapping emotional states. Personalized long-

term support will be enabled through improved

context-aware response generation and emotion trend

analysis. Clinically validated psychological

strategies such as cognitive behavioural therapy will

improve mental health interventions. Multilanguage

and culturally adapted chatbot capabilities will

ensure inclusivity. Privacy preserving techniques

such as federated learning will enhance the security

of the data. Finally real-time alerts are expected to be

put into place as a means for the chatbot to assist

users in extreme distress.

REFERENCES

Artificial Intelligence Powered Chatbot for Mental

Healthcare based on Sentiment Analysis. 2022 5th

International Conference on Advances in Science and

Technology (ICAST).

Fadhil, A., & Moffatt, K. (2020). The potential of chatbots

for mental health support: A systematic review.

Journal of Medical Systems.

Fitzpatrick, K. K., Darcy, A., & Vierhile, M. (2017).

Delivering Cognitive Behavioural Therapy to Mood and

Anxiety Patients through a Smartphone App: A

Randomized Controlled Trial. Journal of Medical

Internet Research.

Laranjo, L., et al. (2018). Conversational agents in

healthcare: A systematic review. Journal of the

American Medical Association (JAMA).

Tielman, M., & de Vries, L. (2020). Chatbots for mental

health: How artificial intelligence can contribute to

psychological well-being. International Journal of

Human- Computer Interaction.

Lee, J. A., & Choi, J. (2020). Applications of conversational

AI for mental health: A literature review. Journal of

Artificial Intelligence in Medicine.

Miller, C. J., et al. (2020). Artificial Intelligence for mental

health care: A review of opportunities and challenges.

American Journal of Psychiatry.

Lau, A., & Wenzel, S. (2019). The role of chatbots in

improving mental health care delivery: A critical

review. International Journal of Psychiatry and Mental

Health.

Radzi, S., & Khamis, M. (2020). AI and mental health: The

impact of conversational agents on anxiety and

depression. Journal of Healthcare Engineering.

Gaffney, H., & Kuss, D. (2021). The role of AI chatbots in

mental health support: A scoping review. Journal of

Behavioural Health.

Ly, K. H., & Andersson, G. (2020). Chatbots and mental

health: A review of the literature. Psychology and

Psychotherapy: Theory, Research & Practice.

Weitzman, E. R., & Parikh, R. (2019). Digital tools for

mental health care: Chatbots, digital interventions, and

more. Digital Health.

Li, Y., & Wang, T. (2020). Exploring the potential of

chatbot-based interventions for mental health. Journal

of Clinical Psychology.

AI-Driven Chatbot for Mental Health Support Using Sentiment Analysis

297