HealthBot

Inspiration

This project is inspired by the surge of need for medical information during the Covid-19 Pandemic. During times when uncertainty and chaos loomed over all aspects of life, threatened by the spread and impact of a global pandemic, people sought a better understanding of Covid-19. In fact, the overall need for common healthcare information also skyrocketed. Amidst rising misinformation and the influx of various sources, both reliable and unreliable, this project emerged as an answer to the uncertain situation. With the aim of providing users with a quick, reliable response - supported by trusted sources - to inquiries about common diseases and guidelines for their diagnosis and treatment, the HealthBot - a healthcare chatbot - was introduced. In the context of Purdue University, we took inspiration from Purdue University Healthcare's Service to provide students with a quick and responsive source of medical advice.

What it does

The HealthBot takes symptoms input by users and returns a basic diagnosis of possible medical conditions based on the symptoms provided (with relevant disclaimers). The Healthbot then provides additional details for a more in-depth diagnosis, and possible remedies and over-the-counter treatments for them.

How we built it

We researched on a few different models and technologies to build an AI chatbot and ended up using Google DialogueFlow for our main HealthBot. After that, we implement it into our website, where we design and customize it with a few more options

Challenges we ran into

The main challenge we ran into was trying to identify the best way to integrate the chatbot from DialogFlow to our source code on the website because Google DialogFlow has many restrictions on the demo version. In addition, we also looked into building a chatbot from scratch by implementing a simple ANN but it didn't seem to work really well as the mechanisms behind this approach is not strong enough for the chatbot to correlate users' variations in patterns with the recorded patterns in the chatbot's training data.

Accomplishments that we're proud of

We're proud that we have managed to learn more about building a chatbot and make a well-designed website to display our first-ever HealthBot.

What we learned

We learned about the techniques behind creating a chatbot - which involves creating an intent that has possible patterns of questions and possible responses from the chatbot. Through this, we get to know the limitation of the current chatbot as the current NLP model for chatbot to learn still needs to a lot more development to achieve better fluency and flow in conversations with humans.

What's next for HealthBot

In order to develop the HealthBot, we can attempt to compile a stronger database in the json file for the chatbot's neural network model to learn more patterns and do a better job in recognizing and identifying users' intents.