🔁 QA Synth Pro v2.0 — Structural Translation Self-Review Template

🎯 Intended Audience
Translators, translation trainees, educators, RLHF data curators

This repository provides a structured and retrainable self-review template for translation quality assurance, focusing on sentence-level structural fidelity and AI-alignment suitability.

---

📐 Evaluation Axes (5D + Optional)

1. Semantic Layout – Semantic core retention, syntax node mapping
2. Structural Flow – Syntactic trunk stability, punctuation, conjunction usage
3. Spacetime Dynamics – Temporal, causal, and emotional sequencing
4. FPE Standards – Post-editing corrections, style appropriateness, register consistency
5. Neural Alignment – Syntactic reproducibility, parallel corpus suitability

💭 Optional: Translation Resonance – Rhythm, imagery, poetic cadence

---

📘 Documentation

• docs/system_instructions.md → Full specification (8000 chars, transparency & reproducibility)
• docs/scoring_logic.md → Pseudo BLEU/ROUGE structural scoring logic
• docs/self_review_axes.md → Five-axis evaluation guide
• docs/examples.md → Sample reviews with revision proposals

---

🧪 Examples

• examples/en-ja_review.md → English→Japanese review case
• examples/cn-ja_review.md → Chinese→Japanese review case
• examples/en-fr_review.md → English→French review case

---

📂 Repository Structure

• docs/ → Full specifications & guides
• examples/ → Case studies (EN→JA, CN→JA, EN→FR, …)
• templates/ → Self-review forms for Notion, Google Docs, Word (planned)

---

⚙️ Deployment

✅ Usable as a self-review sheet in Notion / Google Docs / Word
✅ Compatible with BLEU/ROUGE auto-scores (comparative comments)
✅ Exportable to JSON/CSV for RLHF training and translation QA education

---

🚀 Usage

Example self-review workflow:

# Evaluate a translation draft with pseudo scoring
python evaluate.py --input my_translation.txt --axes 5 --output review.json