Causalitree

extension pop-up
landing page
tree leaf
middle of the tree
glimpse into the future

Inspiration

We are incredibly passionate about events that surround us, and always curious to understand how they developed. We also find that standalone news articles often lack context, leading to misinterpretation and misconception.

What it does

Causalitree provides a clean, friendly, and easy-to-use interface that allows the user to interactively explore the chain of events leading up to a given event through a tree of in-context news articles. The Causalitree can be endlessly traversed, allowing the user to dive deep into any rabbit hole they wish. Causalitree also leverages the chain of events leading up to an article alongside the power of LLMs to generate a creative and relevant prediction for the future to help the user understand where the world is headed!

How we built it

We combined the well-worn technologies of web-scraping using Python's "requests" and "beautiful soup" libraries, alongside a custom google search engine and OpenAI GPT models to generate dynamic and accurate Causalitrees. This was then paired with a flask backend and react frontend to create a clean, responsive, and adaptive web application that is accessible to anyone.

Challenges we ran into

We did not expect to run into so many issues with our web scraping, between being unable to directly scrape urls from "news.google.com", to struggling to extract publication dates from articles that don't contain them. We also struggled to decide on which LLM was most appropriate for our application.