| name | register-translator |
| description | Generate register-conversion rules that map the author's spoken or Chinese-X expressions to compliant academic-English equivalents, preserving the underlying function. Trigger when the user wants to translate voice/tone across registers, build a conversion table, or find academic equivalents for casual/colloquial fingerprint markers. |
register-translator
LAYER 2 · SKILL · table-driven deterministic conversion. The operational annex to voice-synthesizer.
Purpose
Take an author fingerprint expressed in a casual register (spoken / Chinese-X / email) and produce its academic-English equivalent — converting the SURFACE FORM while preserving the UNDERLYING FUNCTION.
Core principle
Cross-register transfer rides on function, not surface form. For every input expression, first name the underlying cognitive/rhetorical function, THEN find the academic-English realization of that same function.
Output: a conversion rule table
Columns: source expression | register | underlying function | academic-English equivalent(s)
Seed rules (extend from real fingerprint data)
| source | function | academic-English equivalent |
|---|
| "we're on the same page" | establish shared premise | "It is uncontroversial that…" / "Taking X as given,…" |
| "I just finished X" | stage-progress self-report | "Having established X in §N, this section turns to…" |
| "Is it correct?" (closing a thought) | make thinking public | "Whether this characterization holds is examined below." |
| "the fucking X" (argument compression) | strong impatience + already-traversed reasoning | "the objection that X — [restate] — rests on a category error" |
Boundaries
- Some fingerprints have NO academic equivalent and should be dropped (e.g. profanity) — but record the underlying function so the academic substitute can carry it.
- This skill produces RULES, not rewritten drafts. Application happens in
tone-auditor.
Self-assessment for scripting
The conversion table itself is data; lookup/application against a draft could be scripted (regex/pattern match → suggest equivalent) once the table is mature. Generating new rules stays LLM-driven (requires function inference). Follow SCRIPTING-POLICY.md if you script the lookup.
