content-locale-humanize-ar · diff
git:20260918.40e721d to git:20261013.a69e0bb
19 added, 7 removed. Audit A to A.
---
name: content-locale-humanize-ar
title: "Arabic Native-Fluency Calibration"
description: "Per-language calibration anchors for detecting AI-slop and translationese in Arabic (ar) target text; auto-loads alongside content-locale-humanize when the target language is Arabic."
license: Apache-2.0
- compatibility: "Extends content-locale-humanize's AI-Slop & Translationese dimension. Load both together — this file has no rubric of its own."
+ compatibility: "Calibration data only; no tools required. Network access helps for usage checks."
domains: content
rules:
- content(arabic)
- match(\bar-(SA|EG|AE)\b)
- match(\bArabic\b)
- match(العربية)
- semantic(check if this Arabic translation sounds native)
- semantic(eliminate translationese from this Arabic text)
---
## Overview
- Sourced calibration anchors for Arabic, feeding `content-locale-humanize`'s AI-Slop & Translationese dimension. This is a calibration aid, not the checklist — reason natively beyond it (see the core skill's "why structure, not word lists" section).
+ Calibration anchors and current norms for Arabic (Modern Standard Arabic unless the brief names a dialect), feeding content-locale-humanize's dimensions. A calibration aid, not the checklist; reason natively beyond it.
- ## Instructions
+ ## Rules
- Sourcing confidence: the weakest of any language checked. Most published Arabic AI-word lists are direct translations of the English canon, not independently observed findings — treat any finding here as a starting hypothesis, not a confirmed tell.
+ ### AI tells
- Stock phrases: "في الختام" ("in conclusion"), "لا شك أن" ("there's no doubt that").
+ Sourcing confidence: the weakest of the set. Most published Arabic AI-word lists translate the English canon rather than report independent findings; treat any item as a hypothesis.
- Beyond this thin list, lean harder on the structural checks in `content-locale-humanize` (interference, connector density, rhythm clustering) than on vocabulary matching — the vocabulary evidence for Arabic specifically is not solid.
+ - Stock phrases: "في الختام" (in conclusion), "لا شك أن" (there's no doubt that).
+ - Lean harder on the structural checks (interference, connector density, rhythm clustering) than on vocabulary matching.
+ ### Translationese from English
+
+ - "يتم + مصدر" passives ("يتم تحديث البيانات") where a native writer uses the internal passive or active voice ("تُحدَّث البيانات").
+ - "من قِبَل" as a calque of the English agent "by"; dummy "قام بـ" + noun; "حيث" as an all-purpose connector.
+ - English word order and stacked pronouns; Latin punctuation (, ; ?) instead of ، ؛ ؟.
+
+ ### Current norms
+
+ - Register: MSA for written product, documentation, and formal copy; Gulf, Egyptian, or Levantine dialect only when the brief targets that market and channel.
+ - Digits: Western (0–9) or Arabic-Indic (٠–٩) per market and house style; hold one choice. RTL layout with isolated LTR runs for code, URLs, and Latin brand names.
+
## References
- One Al Jazeera piece naming a handful of phrases; otherwise mostly translated-from-English lists.
+ Reviewed 2026-10-12. AI tells: one Al Jazeera piece naming a handful of phrases; otherwise translated-from-English lists. Norms: Unicode CLDR locale data; Arabic translation-criticism literature on English interference.