Daeseon Yoo
Back to project
·Update·2 min

Preposition spotlight: teaching the why, not the gloss

A new learning feature: for every meaningful preposition in a clip, the AI explains the relationship it encodes (off of = source you lean on; into = transformation) in Korean — built to fix the habit of guessing prepositions by mapping Korean particles word-by-word.

The pain that prompted this is universal for Korean English speakers: you guess prepositions. off of, into, up with, through — you map them to a Korean particle and hope. A dictionary tells you off = ~에서, which is exactly the gloss that doesn't help, because prepositions aren't words, they're relationships.

So the analysis pipeline now produces a fourth learning lens alongside the existing 직독직해 (chunked translation), key expressions, and practice scenarios: preposition_notes. For each preposition worth explaining in a clip, the AI returns:

The prompt is explicit that a dictionary gloss is a failure ("off of = ~에서" is rejected in the examples); it has to be the why. into isn't "~로", it's "transformation — A reshaped to become B," which is why you refactor a class into modules but not to modules.

The build

It's a thin vertical slice on top of the existing provider-neutral analysis:

Backend suite green, frontend production build green. No new product category — this is squarely the existing "in-app AI explanation" value (NORTH STAR §0.3.3), just aimed at the one thing the owner most wanted to stop guessing.

Next: a fill-in drill that blanks these prepositions for active recall, and a chunk-by-chunk shadowing mode for the "speak in breaths, not word-by-word" half of the problem.