EOC Analyzer

EOC documents can be complex for consumers to understand. Our system extracts relevant info, provides an UI to help users with healthcare coverage, turning 100-page docs into short insights.

Comment

Inspiration

Students and scholars often find themselves having to review lengthy documents filled with a bunch of jargons, and despite the similar nature of some of them (such as slides from the same class), it can be a very daunting task to have to go through all of them. This project helps solve this problem.

What it does

EOC Analyzer take as input an Evidence of Coverage document in PDF form. It then uses multiple method of enrichment, adding contexts and relations to the documents. To name a few of these methods: Dictionaries, Patterns, Extract Entities, and Text Classifier. Instead of the time-consuming plain-text search, this paradigm allows fast querying and understanding of information. Furthermore, a trained model that based on one type of document can also effectively analyze other documents of the same type.

How we built it

The 2 main supporting architecture for this project were Watsonx Discovery, used for Document Retrieval, and Watsonx Assistant, used for conversational communication with AI. I used Discovery to train a model on a collection of EOC document, then utilized that same model to back Assistant, providing users with insights on their Health Insurance Info

Challenges we ran into

Working on this project was a fascinating experience. However, there are a few things that would have made it better. First, I should have collab with other developers to reduce the workload. Secondly, I should have changed my approach of resolving all issues at once, and instead aim for releasing an MVP.

Accomplishments that we're proud of

I feel very happy that I spent all working hours of this hackathon learning about Watsonx AI, as it is a tool that interests me a lot due to how powerful it is and how fun it is to use. Also, despite not having enough time, I got to meet a lot of cool people and help other developers on their projects.

What we learned

After today, I can confidently use the toolkit offered by IBM Watsonx. I can utilize API, train models, and make any apps I want, as long as I have my free trial.

What's next for EOC Analyzer

For the next steps, I have 3 directions:

  1. Also train models for other categories of document, apart from EOC
  2. Improve model using multiple enrichment methods
  3. Improve UI and make a chatbox conversation instead of just a QA prompt

Built With

Share this project:

Updates