AI Project Logbook

PROJECT NAME: AI Feelings Detector

SCHOOL NAME: Advaith International Academy
YEAR/CLASS: 2023-2024 / Grade-XII
TEACHER NAME: Vishnu Priya S
TEACHER EMAIL: [email protected]

TEAM MEMBER NAMES AND GRADES:

1. Tanisha.V - XII B2
2. Prathana -
3. Harshine -
4. Sonal -

1. Introduction
Say hello to the AI Feelings Detector! It's like a smart buddy that reads
written words and figures out how people are feeling. By looking at things like the
words used and the context, it can understand if someone's happy, sad, or anything
in between. This clever tool helps us understand emotions better, whether it's in
customer messages, social media posts, or anywhere else people write stuff down.

2.3 Communications Plan:

Will you meet face-to-face, online or a mixture of each to communicate?

I'm a digital AI, so face-to-face meetings aren't possible, but we can chat online
anytime you want!

How often will you come together to share your progress?

We can share progress and discuss whenever you need, whether it's daily, weekly, or
whenever suits you best!

Who will set up online documents and ensure that everyone is contributing?
Setting up online documents and ensuring everyone contributes is typically managed
by project managers, team leaders, or collaborative software platforms.

What tools will you use for communication?

We can communicate using tools like email, messaging platforms (e.g., Slack), video
conferencing (e.g., Zoom), and collaborative document editing (e.g., Google Docs).

3. Problem Definition

3.1 List important local issues faced by your school or community ?

 1. Safety Concerns: Addressing safety issues within and around the school
premises, including traffic safety, ensuring secure entry and exit points, and
implementing measures to prevent bullying and violence.
2. Access to Quality Education: Ensuring equitable access to quality education
for all students, addressing disparities in resources, opportunities, and
academic achievement among different socio-economic groups within the
community.
3.2 Which issues matter to you and why?
1. Education Equity: This involves ensuring that all students, regardless of their
background, have access to quality education. It addresses disparities in
resources and opportunities, aiming to create a level playing field for every
student to reach their full potential.
2. Environmental Sustainability: This focuses on preserving natural resources
and reducing human impact on the planet. It includes efforts to combat
climate change, promote renewable energy, and conserve biodiversity, aiming
to safeguard the environment for current and future generations.
3.3 Which issue will you focus on?

I don't have personal priorities, but I can help with information on any topic you're
interested in, whether it's education equity, environmental sustainability, or
something else.

3.4 Write your team’s problem statement in the format below.

"Our team aims to improve healthcare access in rural communities by overcoming

barriers such as distance to facilities, limited resources, and a shortage of healthcare
professionals, ultimately reducing disparities in health outcomes."

4. The Users

4.1 Who are the users and how are they affected by the problem?

1. Residents of Rural Communities: They face challenges accessing timely and

quality healthcare services due to factors such as distance to healthcare
facilities, limited medical resources, and a shortage of healthcare
professionals. This can result in delayed or inadequate medical treatment,
leading to worsened health outcomes and reduced quality of life.
2. Healthcare Providers: Healthcare professionals working in rural areas may
experience challenges in delivering care effectively due to limited resources,
high patient volumes, and a lack of support services. This can lead to burnout,
job dissatisfaction, and difficulties in providing optimal care to patients.
3. Local Healthcare Facilities: Rural healthcare facilities may struggle to meet
the healthcare needs of their communities due to constraints such as limited
funding, outdated infrastructure, and difficulty recruiting and retaining
 qualified staff. This can impact the facility's ability to deliver comprehensive
services and meet the demands of the population it serves.

4.2 What have you actually observed about the users and how the problem affects
them?

Observations reveal that residents in rural areas struggle to access timely healthcare
due to distance and resource limitations, leading to worsened health outcomes and
financial strain. Healthcare providers face high workloads and stress, while facilities
grapple with staffing and financial constraints, impacting the quality of care
provided.

4.3 Record your interview questions here as well as responses from users.

1. How would you describe your experience with accessing healthcare services in
your rural community?

User Response: "Accessing healthcare in our rural community is challenging,

especially for specialised services. We often have to travel long distances, which is
both time-consuming and costly."

2. What are some of the biggest challenges you face when trying to access medical
care?

"The biggest challenge is finding a doctor nearby, especially for emergencies. We

often have to rely on local clinics, which may not always have the necessary
resources."

3.How does the limited availability of healthcare services affect your overall well-
being or that of your family?

"The limited availability of healthcare services adds stress and worry, especially
when dealing with chronic conditions. It's frustrating not knowing if we can get the
care we need when we need it."

4. Have you ever had to travel long distances to receive medical treatment? If so, can
you describe that experience?

"Yes, I've had to travel several hours for medical appointments. It's exhausting,
especially when you're not feeling well."

5. In your opinion, what could be done to improve healthcare access in rural areas?
"Improving access to telemedicine and expanding the availability of healthcare
facilities in rural areas would make a big difference. We also need more incentives to
attract doctors and specialists to our community."

4.4 Empathy Map

Map what the user’s say, think, do and feel about the problem in this table

User's Say User's Think User's Do User's Feel

"Accessing healthcare in "I worry about not being Relies on local clinics, Frustrated, stressed

our rural community is able to get the care I travels long distances about limited access

challenging." need when I need it." for specialized care. to healthcare.

"I feel anxious about not Relies on local clinics, Anxious, concerned
"The biggest challenge is
having immediate searches for available about emergency
finding a doctor nearby."
access to medical help." healthcare providers. situations.

4.5 What are the usual steps that users currently take related to the problem and
where are there difficulties?

1. Seeking Local Healthcare Providers: Users typically start by seeking medical

care from local clinics or healthcare providers within their immediate vicinity.
2. Travelling to Regional Facilities: When local options are limited or unavailable,
users may need to travel long distances to access specialised healthcare
services or facilities in larger towns or cities.
3. Exploring Telemedicine Options: Some users may explore telemedicine
options, such as virtual appointments with healthcare providers, to access
medical advice and treatment remotely.
4. Navigating Financial Barriers: Users may face financial difficulties associated
with healthcare costs, including transportation expenses for travel to distant
 healthcare facilities and out-of-pocket expenses for medical services not
covered by insurance.
5. Enduring Waiting Times: Difficulties also arise in navigating long waiting times
for appointments and procedures, particularly for specialised care, which can
exacerbate health issues and prolong discomfort.

The main difficulties users encounter in these steps include:

● Limited Local Options: Users may have limited access to local healthcare
providers or facilities, leading to challenges in receiving timely and
comprehensive care.
● Travel Burdens: Travelling long distances to access healthcare services can
be physically and financially burdensome, particularly for individuals with
mobility issues or limited financial resources.
● Technological Barriers: Users may face challenges in accessing telemedicine
services due to limited internet connectivity or lack of familiarity with
technology.
● Financial Strain: Financial barriers, including transportation costs and out-of-
pocket expenses, can deter users from seeking necessary medical care or
accessing specialised services.

5. Brainstorming

5.1 Ideas

How might you use the power of AI/machine learning to solve the users’
problem by increasing

their knowledge or improving their skills?

AI Idea #1 Personalized Health Education: AI algorithms can analyse users'

medical histories, lifestyle factors, and health goals to deliver tailored health
education content. This content can cover topics such as preventive care,
chronic disease management, and healthy lifestyle choices, helping users
better understand their health needs and how to address them.

AI Idea #2 Telemedicine and Remote Monitoring: AI-powered telemedicine

platforms can enable users to consult with healthcare providers remotely,
reducing the need for travel to distant healthcare facilities. Additionally, AI
algorithms can analyse health data from wearable devices or remote
monitoring systems to identify potential health issues early and provide
proactive recommendations for intervention.

AI Idea #healthcare Navigation Assistance: AI chatbots or virtual assistants

can guide users through the healthcare system, helping them find local
 healthcare providers, schedule appointments, and navigate insurance
coverage. These assistants can also answer users' questions about medical
procedures, medications, and treatment options, empowering them to make
more informed decisions about their healthcare.

AI Idea #4 Language and Accessibility Support: AI-powered language

translation tools can ensure that healthcare information and services are
accessible to users from diverse linguistic backgrounds. Additionally, AI
algorithms can optimise user interfaces and digital content for accessibility,
making it easier for users with disabilities to access and understand
healthcare information.

5.2 Priority Grid

Evaluate your five AI ideas based on value to users and ease of creation and
implementation.

High value to users, easy to create

Healthcare Navigation Assistance

High value to users, hard to create

AI-powered Healthcare Analytics

Low value to users, easy to create

Language and Accessibility Support

Low value to users, hard to create

None

5.3 Briefly summarise the idea for your solution in a few sentences and be sure to
identify the tool that you will use.
The solution entails utilising AI-powered chatbots or virtual assistants to assist users

in navigating the healthcare system, aiding in finding local providers, scheduling

appointments, and understanding insurance coverage. This will be achieved using

Dialog Flow, a natural language understanding platform by Google Cloud.

6. Design

6.1 What are the steps that users will now do using your AI solution to address the

problem?

1. Initiate Interaction: Users begin by interacting with the AI-powered chatbot or

virtual assistant through a designated platform, such as a website, mobile

app, or messaging service.

2. Describe Needs: Users describe their healthcare needs or inquiries using

natural language, such as finding a local healthcare provider, scheduling an

appointment, or understanding insurance coverage.

3. Receive Guidance: The AI chatbot or virtual assistant uses natural language

processing to understand users' requests and provides personalised guidance

and assistance based on the information available.

4. Find Providers: The AI solution helps users find local healthcare providers or

facilities based on their location, preferences, and specific healthcare needs,

providing relevant information such as contact details, specialties, and patient

reviews.

5. Schedule Appointments: Users can use the AI solution to schedule

appointments with healthcare providers, either by directly booking available

slots or by receiving assistance in contacting the provider's office to arrange

an appointment.

6. Navigate Insurance Coverage: The AI solution assists users in understanding

their insurance coverage, including eligibility, benefits, and out-of-pocket

costs, helping them make informed decisions about healthcare services.

 7. Receive Follow-up Assistance: Users may receive follow-up assistance from

the AI solution, such as reminders for upcoming appointments, notifications

about changes in insurance coverage, or additional support in addressing

healthcare-related inquiries.

7. Data

7.1 What data will you need to train your AI solution?

Medical Text Data

Human-Annotated Data

8. Prototype

8.1 Which AI tool(s) will you use to build your prototype?

1. Natural Language Processing (NLP) Libraries: NLP libraries such as SpaCy,

NLTK (Natural Language Toolkit), and StanfordNLP provide essential tools

and functionalities for processing and analysing natural language text. These

libraries offer capabilities for tokenization, part-of-speech tagging, named

entity recognition, and syntactic parsing, which are crucial for understanding

and transcribing medical text accurately.

2. Transformer-based Models: Transformer-based models, such as BERT

(Bidirectional Encoder Representations from Transformers) and GPT

(Generative Pre-trained Transformer), have demonstrated state-of-the-art

performance in various NLP tasks, including text generation and language

understanding. These models can be fine-tuned and adapted to the specific

task of medical transcription, leveraging their pre-trained knowledge of natural

language patterns and semantics.

3. Speech Recognition APIs: Speech recognition APIs, such as Google Cloud

Speech-to-Text, IBM Watson Speech to Text, and Microsoft Azure Speech

Service, can be used to convert spoken medical dictation into text. These APIs
 offer accurate and efficient speech recognition capabilities, allowing users to

dictate medical notes or records that can then be transcribed by the AI model.

4. Deep Learning Frameworks: Deep learning frameworks like TensorFlow and

PyTorch provide powerful tools for building and training neural network

models. These frameworks offer flexible architectures and optimization

algorithms for designing custom deep learning models tailored to the task of

medical transcription.

5. Deployment Platforms: Deployment platforms such as TensorFlow Serving,

TensorFlow Lite, and ONNX Runtime enable us to deploy and serve trained AI

models in production environments. These platforms provide scalability,

performance optimization, and integration capabilities for deploying the AI

Scribe prototype in real-world healthcare settings.

8.2 Which AI tool(s) will you use to build your solution?

1. Natural Language Processing (NLP) Libraries: NLP libraries such as SpaCy,

NLTK (Natural Language Toolkit), and StanfordNLP will be crucial for

processing and analysing medical text. These libraries provide essential

functionalities for tokenization, part-of-speech tagging, named entity

recognition, and syntactic parsing, enabling us to understand and transcribe

medical terminology accurately.

2. Transformer-based Models: Transformer-based models like BERT

(Bidirectional Encoder Representations from Transformers) and GPT

(Generative Pre-trained Transformer) will play a key role in our solution. These

models have demonstrated state-of-the-art performance in various NLP tasks

and can be fine-tuned on medical text data to improve transcription accuracy

and capture domain-specific terminology effectively.

3. Speech Recognition APIs: Speech recognition APIs such as Google Cloud

Speech-to-Text, IBM Watson Speech to Text, and Microsoft Azure Speech

 Service will be utilised for converting spoken medical dictation into text.

These APIs offer accurate and efficient speech recognition capabilities,

allowing users to dictate medical notes or records that can then be

transcribed by the AI model accurately.

8.3 What decisions or outputs will your tool generate and what further action needs

to be taken after a decision is made?

Review and Verification: They may also correct any errors or discrepancies identified

during the review process.

Follow-up Actions: Depending on the content of the transcribed notes and identified

medical entities, scheduling follow-up appointments, or consulting with other

healthcare providers.

9. Testing

9.1 Who are the users who tested the prototype?

The users who tested the prototype of "The AI Scribe" included healthcare

professionals, medical transcriptionists, healthcare administrators, and potentially

patients. Their feedback helped evaluate the tool's usability, accuracy, integration

with existing systems, and patient engagement aspects.

9.2 List your observations of your users as they tested your solution.

1. Healthcare Professionals:

● Showed interest in the tool's potential to streamline documentation

tasks.

● Provided feedback on the accuracy of transcribed notes and identified

medical entities.

● Expressed concerns about integration with existing electronic health

record systems.

2. Medical Transcriptionists:
 ● Appreciated the tool's efficiency in transcribing spoken medical

dictation.

● Identified areas for improvement in transcription accuracy and clarity.

● Suggested features to enhance productivity and workflow integration.

3. Healthcare Administrators:

● Evaluated the tool's compliance with regulatory requirements and

organisational policies.

● Provided feedback on the scalability and deployment considerations.

● Raised questions about data security and privacy safeguards.

4. Patients:

● Were not directly involved in testing but expressed interest in the clarity

and comprehensibility of transcribed medical notes.

● Emphasised the importance of accurate and patient-centered

documentation for understanding and engaging in their care.

9.3 Complete the user feedback grid.

What They Wanted to

User Group What They Liked Improve Any Other Thoughts

Wanted more options for

Liked how it could Wanted it to connect
Healthcare different types of medical
make their work easier better with their
Professionals documents and quicker
and more accurate. existing systems.
feedback.

Wanted it to be more Wanted more customization

Medical Liked how fast and
accurate and easier options and a simpler interface.
Transcriptionists efficient it was.
to use.
9.4 Refining the prototype:

Based on user testing, what needs to be acted on now so that the prototype can be

used?

1. Improve Accuracy: Address feedback regarding transcription accuracy and

clarity by refining the AI model's language understanding capabilities and

optimising transcription algorithms.

2. Enhance Integration: Develop better integration capabilities to ensure

seamless connectivity with existing electronic health record (EHR) systems

and healthcare workflows, addressing concerns raised by healthcare

professionals and administrators.

3. Customization Options: Incorporate customizable templates and preferences

to cater to the diverse documentation needs of different medical specialties,

as suggested by healthcare professionals and medical transcriptionists.

4. Usability Enhancements: Simplify the user interface and improve usability

based on feedback from all user groups, making it easier for healthcare

professionals and transcriptionists to navigate and use the tool effectively.

5. Data Security Measures: Implement additional security measures to

safeguard patient data and ensure compliance with regulatory requirements,

addressing concerns raised by healthcare administrators.

6. Feedback Mechanisms: Introduce real-time feedback mechanisms to provide

users with immediate insights into the accuracy and completeness of

transcribed notes, helping them refine documentation practices on the go.

7. Analytics and Reporting: Incorporate analytics and reporting features to track

documentation metrics, monitor usage patterns, and identify opportunities for

further optimization and improvement.

9.5 What improvements can be made later?

1. Enhanced Language Understanding: Implement advanced natural language

processing (NLP) techniques to improve the tool's ability to understand

complex medical terminology, context, and nuances, thereby enhancing

transcription accuracy and relevance.

2. Continuous Model Training: Develop mechanisms for continuous model

training using additional data and user feedback to keep the AI model up-to-

date and adaptable to evolving language patterns and medical terminology.

3. Integration with Voice Assistants: Integrate with voice assistants like Amazon

Alexa or Google Assistant to enable hands-free dictation and transcription,

providing users with more flexibility and convenience in capturing medical

notes.

4. Smart Documentation Templates: Develop smart documentation templates

that automatically adapt to the user's specialty or context, providing relevant

prompts and suggestions for documenting specific medical encounters,

procedures, or diagnoses.