DocBot

• 

  launch page

• 

  Joe Biden trying out our app

• 

  A possible diagnosis

Inspiration

Users today are increasingly averse to storing personal data on their phones. However, we need data-driven functionality to maintain a healthy, growing society. We were inspired to use top-of-the-line chatbot architecture and natural language processing to implement a seamless, unintrusive service: an AI doctor.

We saw practical implications, ranging from third-world awareness to first-world convenience. We were motivated to provide a service that is beautiful and intuitive, yet provides fast and appropriate diagnoses to users.

What it does

DocBot does just what sounds like; it is a self-learning "bot" that analyses users' health through conversation. It asks simple questions -- what your name is, how you are feeling today, etc -- and uses your answers to diagnose your ailments. DocBot recognizes a variety of symptoms from commonly occurring illnesses, ranging from strep throat to heart attacks.

How we built it

Coming to MHacks, we hoped to create a simple but relevant Machine Learning application. DocBot's front-end was created using MEAN stack -- AngularJS on the front-end, MongoDB, ExpressJS, and NodeJS on the backend. We also used Microsoft's ChatBot framework as well as their LUIS natural language processing API. We connected our bot with Twilio to give access to non-smartphone owners.

Challenges we ran into

We came to MHacks with very little, almost no, machine learning experience. A substantial amount of time was devoted to deciphering the concepts and the implementations we wanted. The biggest challenge we ran into, however, was coordination. We all wanted to dive head-first into a machine learning application, but had to keep our heads and delegate work to finish on time.

Accomplishments that we're proud of

We're proud to have a working, learning machine learning algorithm that recognizes symptoms. We also are proud of our styling -- the application looks sleek and professional.

What we learned

We learned how to appropriately delegate responsibilities and were able to learn both as a team and by ourselves. We also gained insight into the process of a machine learning algorithm.

What's next for DocBot

We hope to expand the symptoms and diseases DocBot recognizes. We also want to connect doctors to our chatbot, providing a portal for them to train the neural network.

Built With

• angular.js
• bootstrap
• express.js
• git
• javascript
• mongodb
• node.js
• python