We were inspired by the fact that a TL:DR is included at the start/end of most news articles, but there was no TL:DR for parliament hearings. News outlets are a great way of keeping track of what's going on in Canada, but news mostly consists of mainstream headlines and is not at all inclusive of all decisions that are being made in our country. We saw the opportunity, and went for it! A TL:DR for parliament hearings!

What it does

Our app scrapes proceedings from the House of Commons of Canada and runs it through the Article Summarization feature of Cohere API.

Once summarized, it roughly interprets the topic of the hearing and passes it as a prompt to Wombo API, fetching an AI-generated image, as a small entertainment factor for the user.

For scalability, it also stores summaries and AI-images for each proceeding in a database, so that if N users request to see the summary of a proceeding, it does not have to scrape and summarize N times, but only one time. After that, it will fetch the same result from the database.

How we built it

1) Web scraper runs through parliament hearing transcripts on House of Commons website 2) Scraped article is passed to Cohere API for article summarization 3) Topic of parliament proceeding is interpreted and passed to Wombo API to generate an image 4) Image and summary is returned to user

Challenges we ran into

We had to play around with and fine tune the Cohere API hyperparameters to achieve optimal results, which consumed a lot of time

Accomplishments that we're proud of

Scraping random samples from parliament website Connecting everything together after implementing modular functions for each piece of the puzzle Successfully summarizing parliamentary proceedings and providing a useful result

What we learned

That the Cohere API is a very comprehensive and capable API, with impressive training data

What's next for Co:ngress

Drawing a cloud infrastructure diagram for production deployment that will support up to 36 million users :)

Built With

Share this project: