boston-translatahlisted

# Boston Translator Turn any text Boston. Two engines run in order; a **thickness dial** controls how hard they push; a **regional mode** picks the vocabulary and attitude. > This skill is also a teaching artifact. The whole point of the repo is to give > non-developers a low-stakes way to learn skill files, markdown, and pull > requests. If you're reading this to *contribute* rather than run it, go to > `CONTRIBUTING.md`. ## Data model — three layers 1. **Pronunciation rules** — `data/pronunciation/rules.yml` General phonetic transforms that apply to *any* word (drop-R, intrusive-R, broad-A, o→aw, -er→-ah). Curated and small. This is the accent engine. 2. **Lexicon** — `data/lexicon/<region>/<slug>.yml` One file per term. Standard word/phrase → Boston equivalent. The big, community-grown dictionary. This is where PRs land. 3. **Phrases** — `data/phrases/<region>/<slug>.yml` Canonical multi-word renderings that do NOT derive cleanly from the rules (irregular idioms like "pahk the cah in Hahvad Yahd"). Stored verbatim. Every lexicon and phrase entry carries a `sources:` block. No source, no merge. ## The thickness dial | Level | Name | What it does | Use for | |---|---|---|---| | 1 | **Light** | Lexicon swaps only, register-1 terms (wicked, packie, Dunks). No accent respelling. | Real content you'd actually publish. Still reads clean. | | 2 | **Local** | Register 1–2 swaps + soft accent (drop trailing R, -er→-ah). | Clearly Boston, still readable. | |