nusantara-corpus-pdf-extractor
Installation
SKILL.md
Nusantara Corpus PDF Extractor
Turns a scanned or digital dictionary (local language ⇄ some pivot/gloss language) into a quality-checked JSONL parallel corpus, through a loop, not a single pass.
Fully language-agnostic by design: which local language, which pivot language (Bahasa Indonesia by default, or whatever the dictionary glosses into), and which orthography rules apply are all supplied as config — nothing in the pipeline logic assumes a specific language pair. Originally generalized from Lani dictionary extraction work, but "Lani-specific" behavior has been deliberately designed out.