Inspiration
YouTube, the world’s largest video-sharing site has grown tremendously as a video content platform, with the recent shift in online content to video.
YouTube creators can monetize if they have more than 1,000 subscribers and 4,000 hours of watch time for the last 12 months. Accordingly, spam comments are being created to promote their channels or videos in popular videos. Some creators closed the comment function due to aggression such as political comments, abusive speech, or derogatory comments unrelated to their videos.
YouTube has its own spam filtering system, though there are still spam comments that are not being caught. So there is a need for a rigorous system that gives more control to the content creators in spam control.
What it does
We hence built a command line app (Retracto) that lets you easily identify spam comments in a video and delete them.
How we built it
I experimented with the Cohere platform and fine-tuned a model with the dataset https://www.kaggle.com/code/paliwal1/youtube-comment-classification/data from Kaggle.
Architectural Diagram
Challenges we ran into
I faced a lot of challenges around the Google OAuth and Browser SSO of the Command Line App. I had to research on how to open the browser and listen to the redirect of OAuth
What's next for retracto
Retracto can be extended to include auto-reply to trivial comments and appreciations. This can save a lot of time for content creators plus it will keep the users engaged.
Log in or sign up for Devpost to join the conversation.