←back to thread

456 points ph4evers | 2 comments | | HN request time: 0.397s | source

I've been working on a little side project that combines Duolingo-like listening comprehension exercises with real content .

Every video is transcribed to get much better transcripts than the closed captions. I filter on high quality transcripts, and afterwards a LLM selects only plausible segments for the exercises. This seems to work well for quality control and seems to be reliable enough for these short exercises.

Would love your thoughts!

1. iamkonstantin ◴[] No.43543968[source]
Did you hand-pick the videos? My first one was some Elon Musk conspiracy dumpster and the second one some church “morality” thing… I think it’s a good example of what not to do with LLMs.

Also, your page needs to disclose any content filtered by or generated by a model.

replies(1): >>43544371 #
2. ph4evers ◴[] No.43544371[source]
No I let the LLM filter on "non-war and non-politics" but I don't have a ton of content available (yet) so it might picked something that was not great. Which language did you try?