FlashLawyer

Inspiration

There are tons of legal document generators on the web (FormSwift, LawDepot, etc.), but they all rely on big clunky multi-step HTML forms. What if a chatbot could replace forms, validation errors, text search, help pages, tooltips, modals, and hyperlinks?

What it does

My vision is a legal document builder that feels like speaking to a human lawyer.

The bot starts by determining the legal document required by the user. For example, a user might say "Selling my boat", then the bot pull up the Boat Bill of Sale template.
It will then initiate a series of questions to fill out the main variables/slots in the form, such as "What is the Boat Make and Model?""
When complete, the bot will automatically save the document, then email or download it at the user's request.

How I built it

Webapp with Angular4, TypeScript, and RxJS. AWS Lambda with Python

Challenges I ran into

Making Lex accept general input. Most legal document data is not iterable, so it's been hard to build a model that will catch all utterances.
Integrating the TypeScript AWS SDK with Angular.
Keeping Lex synced with the application state.

Accomplishments that I'm proud of

I am proud of developing an app that is asynchronous, serverless, and that uses NLP features. This is perhaps the first open source example of a Lex chatbot on Angular and just scratches the surface of future possibilities in web development.

What I learned

I've learned a ton about Lex and how to use NLP to enhance the user experience. I also learned that lambda makes it easy to build specialized services across different platforms and languages.