sustainabot

Inspiration

Are you passionate about sustainability? Are you tired of scrolling through thousands of articles but just can't quite find what you're looking for? Wish you could just talk about what you're looking for and it magically appears right in front of you?

Enter Sustainabot.

What it does

This is Sustainabot. The idea behind Sustainabot is that sustainability begins with a conversation. SustainaBOT is a chatbot that uses natural language processing to understand you. The bot then generates a response based on what it thinks you are asking about and repeats until the inquiry is narrowed down to one specific topic. The bot then returns the best summary of a resource to address your sustainability issues.

How we built it

Front end: We used p5.js to create the chatbot's interface and figma to create the design logo.

Back end: We created a server using python and Flask and hosted it on Google Cloud. The server handles the client's POST requests and gives back a bot response to be displayed. The bot response is built using Cohere, first to classify user responses and then to generate a summary of the most relevant web link.

Challenges we ran into

We ran into issues while building the web scraping function due to the limitations in the model. Feeling that it was time consuming and didn't align with our main goal, we implemented text file stand-ins to focus our efforts on the newer technologies like Cohere's classify.

Accomplishments that we're proud of

We are proud of our group's ability to create a full stack prototype that effectively capitalized on all of the proposed technologies features.

What we learned

We got a lot of experience with Google Cloud Platform's Cloud Run features and learned a natural language processing language like Cohere.

What's next for SustainaBot

Given more time, we would love to create more layers to the conversation and generate an algorithm for the bot to narrow down the search of the topic on its own. We want the bot to be able to scrape the web for search results rather than the text stand-ins we have in place. We would like to implement a speech to text feature for those who have difficulties with typing on a keyboard utilizing AssemblyAI's technologies.