Beebe

• 

  chatbox

Inspiration

I started this project with idea generator plugin and got 3 good ideas based on my skills and interests. I went further to continue using external AI search tool to explore those initial ideas and found more areas where Dropbox Sign and AI could be extremely useful. There were options for project ideas, but I picked the field of intellectual property over ballot casting, delivery, notary systems due to research and genuine interests. I was once recommended to get a book on creative IP for artists and creatives so that lead me to choose this topic to work on.

What it does

Beebe is an IP chatbot web app that enables text chat with a custom trained GPT2 model. It has been developed to to help users understand key trademark terms.

How we ideated it

• Using Flask to route user inputs to the model with a simple frontend.
• Integrating opensource GPT-2 LLM for trademark and IP conversation.

Challenges we ran into

• Finding opensource data for the model.
• Preparing the data in text and json formats using python.
• Getting access to freely available compute resources to train a LLM model.

Accomplishments that we're proud of

• Doing research on Trademark, Intellectual Property for domain knowledge;
• Integrating frontend and generative model with Flask;
• Writing user documentation for Github.

What we learned

• Learning more about product design and how users will interact with prototypes

What's next for Simply Sign

• UI improvements for accessibility tech users
• Updating model with more trademark and IP terms

Built With

• css3
• dropbox-sign
• github
• html5
• javascript