post-edit
c4p0-postedited
c4p0 (post-edited)
Companion to alvations/c4p0. Rows from the upstream consolidated
alvations/c4p0 with a non-empty source AND non-empty translation are
run through Claude Haiku 4.5 to:
Fix encoding issues in the source (mojibake -> Unicode).
Detect the source language.
Clean the noisy LLM translation (strip "Translation:" prefixes,
wrapper quotes, formatting noise) — kept even when the underlying
translation is poor.
Generate a fresh, clean reference translation from the source.… See the full description on the dataset page: https://huggingface.co/datasets/alvations/c4p0-postedited.asr-post-editwmt24-postedit_bw_prefsbhashik-postediting-domain-corporabhashik-errormarked-postediting-domain-corpora