Overview

We decided to focus on the healthcare sector, as access to affordable and immediate health care advice is not an option that everyone has. While googling symptoms is quite common, the results often don't immediately take into account important contributions like the user's past pertinent medical history or co-morbid symptoms like a physician would. This project consists of a chrome extension and a backend NLP-based classifier; the extension directly scrapes a user's medical search from the google search bar and passes it to our backend for processing. From there, the text is broken down into core components and roughly classified into a medical category. The extension then prompts the user to answer some more relevant questions from that category and provides instant, free advice on whether a physician should be consulted. Note: This tool is NOT meant to diagnose conditions or replace the advice of a physician but rather intended to improve the existing 'webMD rabbit hole' process many of us go down when trying to identify what medical conditions we might be experiencing.

Components

The frontend portions were written in javascript by Jorge, Ryan, and Mark and the backend was written in python by Elisabeth.

Frontend

The extension pulls relevant medical-related searches from the search bar and POSTs to the server. Then once a category is determined, the extension asks a series of followup questions to better refine the user's question

Backend

Sentence extracted from the search bar is POSTed using Flask and read into an NPL object. SpaCy is used to tokenize and lemmatize the input data and return the best match to classes of key terms.

UI

We spent the majority of our time developing and connecting the extension and backend components so the time we had left was put towards demonstrating the functionality rather than the aesthetics of the extension. Ideally, we intended to only take up a small portion of the screen with the extension window once the query is activated but alas there are only so many hours in a day.

Future Plans

Integrating patient data through external sources like 23 and Me could be used to better refine pertinent suggestions. In the future, database integration and more robust NLP processing would also be ideal. Connecting/collaborating with a medical provider like Kaiser Permanente and integrating existing direct pipelines to connect with a patient's physician through the extension would also help make the process of acquiring and sharing medical knowledge with your doctor quicker and more intuitive (however securing user data would be a large portion of the process).