Most Complex Translation Prompts?
Soh

Interestingly, it appears that the set of AI/LLM translation prompts I’ve developed and refined over time are among the most complex publicly available today — though I’m sure some people have even longer, customized versions kept private.

I welcome anyone to share if you come across something better.


"Question: are there any other prompts by others more comprehensive than these? https://www.awakeningtoreality.com/2025/04/ai-gemini-prompt-to-translate-atr-blog.html

ChatGPT said:

Short answer: I couldn’t find any public “translation prompts” that are clearly more comprehensive than your ATR Gemini prompts. What exists falls into two buckets:

  1. Prompt libraries/examples — useful, but much simpler than yours

  • Anthropic’s Prompt Library has a basic “Polyglot superpowers” translator template (single-pass, no QA/revision scaffolding). Claude Docs

  • Google’s Gemini Prompt Gallery is broad, but doesn’t include a deep translation workflow (mostly app snippets; no ISO-style steps or bilingual revision). Google AI for Developers

  • OpenAI’s prompting guides (GPT-4.1/GPT-5) give best-practice structures, but they’re general, not domain-specific translation workflows. OpenAI Cookbook+1

  1. Quality standards you can import into a prompt (stronger than most “prompts”)
    If you want to go beyond your current spec, the best “upgrade” is to bake industry standards straight into your prompt’s QA section:

  • ISO 17100 (requires bilingual revision by a second linguist, terminology control, documented workflow). You can translate its steps into checklists inside your PREP/QA stages. ISO+1

  • MQM / TAUS DQF error typology for scoring and error classes (Accuracy, Terminology, Fluency, Style, Locale, etc.). Add an MQM table and require the model to tag each fix with an error type + severity. tranquality.info+1

Bottom line

  • Public prompt templates are simpler than your ATR prompts. None I found ship with your level of segmentation rules, “no-compression” constraints, and multi-pass QA. Claude Docs+2Google AI for Developers+2

  • The only meaningful way to be “more comprehensive” is to embed ISO 17100 process gates and MQM/DQF error-typing into your prompt’s review phase. That gives you an objective QA layer beyond stylistic advice. Iteh Standards+1

If you want, I can draft a compact “ATR Prompt v5.x + ISO 17100 + MQM add-on” block you can paste beneath your current PREP/QA—basically: (a) declare ISO roles & steps, (b) require a bilingual completeness checklist, (c) output an MQM error log with counts and final score."

Labels: | edit post
0 Responses
Subscribe to: Post Comments (Atom)