←back to thread

470 points ph4evers | 1 comments | | HN request time: 0.2s | 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!

Show context
black_puppydog ◴[] No.43544170[source]

Hey, this looks really nice and worked like a breeze for French!

Question: out of the processing steps you mention - transcription, quality filtering, segment selection, and (I guess) wrong-word selection) are there any truly manual steps? Those would be the ones that prevent you from building this for just about any language that has good transcription available, right?

replies(1): >>43544354 #
1. ph4evers ◴[] No.43544354[source]

There are no manual steps. But it is hard to gatekeep quality. The transcription models work well for the large languages but not so much for the smaller ones.