Inspiration
We were inspired by various projects that were designed to interpret news separately from the publisher and make it all available from an app that could be easily placed on a phone or device. This whole idea is to keep our community more updated on current events like town elections or even as large as national events. Since there is a lot of news that isn't backed up or even real, this AI would go in and corroborate sources and show summaries that have had multiple appearances throughout the web.
What it does
Our project starts by gathering and scraping data from the websites and then plugs it into a deep learning machine that we programmed from the Kaggle notebooks here at Kaggle AI This is how we did of our research on all of the Deep learning content from this project, and then it should place it all on a website. Our project is basically an auto-corroborator that will show LLM style prompts for the sources that have multiple appearances.
How we built it
In order to gather the data from the article, we made a scraper which intakes a url and parses an article to return the text. Then, an AI we coded takes that text, and summarizes it. The Kaggle notebooks really helped us focus on making this possible and it helped us doing research on topics like web scraping and AI generation.
Challenges we ran into
We've ran into countless errors, with many issues arising with gathering the article data from the links, then we had many problems with the AI which we hope to fix in the future so that this is more usable.
Accomplishments that we're proud of
We're incredibly proud of our AI interpreter, which uses our multi step approach from taking in the url, to scraping the text, to feeding the AI for its summary, we are really proud of all the new content and topics we learned.
What we learned
Over the course of this development, we've learned so much about handling apis, machine learning, and connecting different programs. We learned a lot about API and how the relationship between the code you use is very important for the project, and you have to keep all the versions up to date. We learned a lot about full stack management.
What's next for Untitled
In our site, you may see placeholder articles named "Recent Topics", in the future, we plan to implement a scraper to independently select articles to recommend, we also plan to finish the AI and make the scraper fully automatic. This could also become an App.
Log in or sign up for Devpost to join the conversation.