Inspiration
Seeing lots of articles and videos, especially on social media, floating around, and wanting a check on whether the writing is prejudiced, since it’s hard to constantly be scanning everything you are looking at.
What it does
BiaScan looks over the inputted text, and scans it over, flagging biased statements, explaining why they are biased, and even offering suggestions on what the biased statement could be changed to.
How we built it
We used an on-board AI with a specific prompt to rate biased phrases, and give explanations why and replacements to them in a JSON format so we could highlight the Biased text and give suggestions
Challenges we ran into
We ran into a few challenges when making this program, namely the highlighting text part, as we previously used character indices rather than looking through the words, which the AI could not count correctly so the highlighting was off
Accomplishments that we're proud of
We are really proud that the model actually works, since this is our first time integrating an AI API into code, and the fact that it can run entirely locally as well is something we are very happy with as well. It is literally a portable checker, like autocorrect, and not too big either, even account for the model, it’s under 2 GB
What we learned
We learned a lot about using AI API, as well as writing python code, especially using the (libraries), and the model integration. Our skills in html also improved, although we had both used it before, it wasn’t to this extent, and creating all the fine details took a while, but it was worth it in the end.
What's next for The Biascan
We plan to try to turn this into a chrome extension, as having it be even more portable, and able to scan web pages would be the most optimal application. A truly portable way to scan any page you want and check for any bias it shows.
Log in or sign up for Devpost to join the conversation.