Educating the population about medical insurance and the complex processes surrounding its usage is especially critical during times of disaster and crisis, such as the Covid-19 pandemic.

Understanding medical insurance specific terminologies involved and choosing the right medical insurance plan (there are thousands of plans by different providers in the US) can be quite difficult, especially for those with limited previous exposure to insurance and those from under resourced communities.

During such times of financial pressure, it is in the interest of medical Insurance companies to harshly enforce certain fineline clauses and deny insurance claims. The enforceable legal clauses and promises are found in the long, legal-speak worded policy documents. We wanted to leverage the power of GPT-3 to convert these contract clauses to human-like simpler versions.

What it does

Automated field extraction and argument mining from Claim Denial Letter + Appeal Letter Generation

Models Used

GPT-3 API provided by OpenAI TF-IDF Text Summarization (also BERT) OCR through GCP Document AI

Challenges I ran into

GPT-3 Token Limit PDF Table Parsing Chrome Extension App Deployment on Google Cloud Platform

Accomplishments that I'm proud of

Making it completely

What I learned

OCR and NLP Techniques

What's next for

Making it scalable

Demo Videos

Detailed Demo Video of all 3 tools part of the dashboard:

Medisure Contract Simplify Chrome Extension (powered by GPT-3) Demo Video:

Share this project: