DOTbot

Ethan Jones, Alyssa Romportl, Keegan Murray, Theresa Worden

Inspiration

One of our team members recently booked a flight for a birthday weekend getaway. Out of curiosity, she headed over to the Department of Transportation's Aviation Consumer Protection website. She browsed the different sections and thought, "With so many people flying these days, you would think this information would be more accessible." She imagined a feature that would answer questions based on Aviation Consumer Protection guidelines in a quick and easy way. Thus, the idea of the DOTbot was born.

What it does

Using information from the Department of Transportation's website, this web application can answer any questions you may have about air travel.

DOT bot uses a generative model specifically trained from a variety of questions and answers from the DOT Aviation Consumer Protection website to produce responses to user input.

How we built it

For this project, we used React's Next.js framework with TailwindCSS and TypeScript to try and build a web app for our project with a modern design. Furthermore, the OpenAI API was integrated in order to train our bot. Originally, we used pythons chatterbot library, but after running into some integration issues with Next.js, we switched to OpenAI's API.

Challenges we ran into

Integrating python code to work in conjunction with our Next.js project was more difficult than planned. Due to last minute difficulties with deploying a work-around, we ended up switching libraries.

Additionally, in the beginning stages of implementing data into the chat bot, we noticed the response time of the bot got longer and longer with an increase of questions/answers added. To speed up the bot's response time, we moved our data into a .json file that increased the bot's reply speed significantly.