Feature Request : Youtube Clickbait Classifier

[alwinraju]

Problem

I often feel frustrated when I open a YouTube video and discover that it lacks value or is merely clickbait, with no relevance to the title or thumbnail that initially caught my attention.

Solution

To address this issue, the extension could be developed to access YouTube transcripts and compare the video's content with its title. By analyzing the match between the two, the extension could classify whether the video is clickbait or not. It would be useful to only run this on demand, i.e make the analysis fire on a button click.

While I'm not entirely certain about the accuracy of the classification process, I believe it could be achievable with a few-shot prompting approach and would be happy to help in curating the prompt.

[MTheSmartestM]

Perhaps you can try doing this yourself to test for the accuracy of the classification process:

1. copy a video's transcript and title
2. put them in chatgpt with a sensible prompt
3. ammend the prompt so that chatgpt only outputs a single word like 'FALSE' for clickbait and 'TRUE' for other vids

4. compare chatGPT's output with your personal opinion of the video (and comeback here perhaps)

   _{psst. I want to clarify that I am not a moderator or developer of the chatGPT extension, but I hope this suggestion helps you test its accuracy. Good luck!}