--- description: Translate the translation back to source language and show side-by-side comparison with drift annotations. argument-hint: " [--unit ] [--report]" --- # /scr:back-translate -- Back-Translation Verification Translate the translation back to the source language and show a side-by-side comparison with drift annotations highlighting meaning changes. This is the primary quality verification tool for translations per D-03. ## Usage ``` /scr:back-translate [--unit ] [--report] ``` - `` -- Target language code to verify (e.g., `fr`, `de`, `ja`, `ar`) - `--unit ` -- Scope to a specific translated unit instead of all units - `--report` -- Generate a standalone back-translation report file ## Instruction You are a **translation verification specialist**. Your job is to perform back-translation -- translating the translated text back to the source language -- and then compare it with the original to detect meaning drift. This is a standard quality assurance technique in professional translation. The back-translation is performed by you (the AI agent) in-context. No external translation API is needed for this step -- you read the translated text and produce a faithful back-translation in the source language. --- ### STEP 1: LOAD CONTEXT 1. Load `.manuscript/config.json` for `source_language` 2. Load Scriveno's installed/shared `CONSTRAINTS.json` (global `~/.scriveno/data/CONSTRAINTS.json` or project `.scriveno/data/CONSTRAINTS.json`) -- check prerequisites: `back-translate` requires `translate` 3. Validate the `` argument is in the `target_languages` list. If not: > Language "[language]" is not in your target languages. Configured target languages: [list]. 4. Check prerequisite: translation exists for this language - Verify `.manuscript/translation/{lang}/drafts/` directory exists and contains files - If no translation found: > No translation found for "[language]". Run `/scr:translate ` first. --- ### STEP 2: LOAD SOURCE AND TRANSLATION For each unit to verify: 1. Load the **original source text** from `.manuscript/drafts/body/{unit}-DRAFT.md` 2. Load the **translated text** from `.manuscript/translation/{lang}/drafts/{unit}-DRAFT.md` 3. If `--unit` is specified, load only that unit pair. Otherwise, load all available unit pairs. If a source file exists but its translation does not (or vice versa), note the mismatch and skip that unit: > **Skipped:** Unit "[unit]" -- [source/translation] file missing. --- ### STEP 3: PERFORM BACK-TRANSLATION For each translated unit: 1. Read the translated text carefully 2. Translate it back to the source language as faithfully as possible - Preserve the meaning of the translation, even if it differs from the original - Do NOT look at the original while back-translating -- translate what the translated text actually says - Use natural phrasing in the source language, not word-for-word rendering 3. Segment both texts into comparable units (paragraphs or meaningful passages) --- ### STEP 4: SIDE-BY-SIDE COMPARISON WITH DRIFT ANNOTATIONS Present a three-column side-by-side comparison per D-03: ``` ## Unit: [unit name] | # | Original | Translation | Back-Translation | Drift | |---|----------|-------------|------------------|-------| | 1 | [original segment] | [translated segment] | [back-translated segment] | [annotation] | | 2 | [original segment] | [translated segment] | [back-translated segment] | [annotation] | ``` **Drift annotation markers:** | Marker | Meaning | When to Apply | |--------|---------|---------------| | **[OK]** | Meaning preserved | Back-translation conveys the same meaning as the original, even if wording differs | | **[DRIFT: meaning shift]** | Semantic change | Back-translation differs from original in factual content, intent, or implication | | **[DRIFT: tone shift]** | Register/emotion changed | Back-translation has different emotional register -- more formal, less urgent, different mood | | **[DRIFT: omission]** | Content missing | Something present in the original is absent in the back-translation | | **[DRIFT: addition]** | Content added | Something in the back-translation was not in the original | **Annotation detail:** For each drift marker, add a brief explanation: - `[DRIFT: meaning shift] -- "courage" became "recklessness", changing the character's motivation` - `[DRIFT: tone shift] -- formal register replaced casual dialogue, losing character voice` - `[DRIFT: omission] -- metaphor about the sea was dropped entirely` - `[DRIFT: addition] -- explanation added that was not in the original text` --- ### STEP 5: SUMMARY STATISTICS After all units are compared, present summary statistics: ``` Back-Translation Summary: [Language Name] ========================================== Units verified: [count] Total segments: [count] Drift Analysis: [OK] -- meaning preserved: [count] ([percentage]%) [DRIFT: meaning shift]: [count] ([percentage]%) [DRIFT: tone shift]: [count] ([percentage]%) [DRIFT: omission]: [count] ([percentage]%) [DRIFT: addition]: [count] ([percentage]%) Overall fidelity: [percentage]% of segments preserved meaning ``` **Fidelity assessment:** - **90%+ OK** -- Excellent translation fidelity. Minor drifts are expected and acceptable. - **75-89% OK** -- Good fidelity with notable drift areas. Review flagged segments. - **Below 75% OK** -- Significant drift detected. Consider re-translating problem units with `/scr:translate --unit `. --- ### STEP 6: REPORT OUTPUT **If `--report` flag is provided:** Write the full report to `.manuscript/translation/{lang}/BACK-TRANSLATION-REPORT.md`: ```markdown # Back-Translation Report: [Language Name] **Source language:** [source_language] **Target language:** [language] **Date:** [current date] **Units verified:** [count] **Overall fidelity:** [percentage]% ## Summary Statistics [Summary table from Step 5] ## Detailed Comparison [Full side-by-side tables from Step 4, organized by unit] ## Recommendations [Based on drift patterns, specific recommendations for revision] ``` **Always display** the summary statistics to the writer, even without `--report`. --- ### OUTPUT Present the summary statistics and top drift issues to the writer. If there are many segments, show only the drift-flagged segments in the terminal output and note that the full comparison is available in the report. ``` Back-translation complete for [Language Name]. Fidelity: [percentage]% ([assessment]) Segments: [OK count] preserved, [drift count] with drift [If drift found]: Top drift issues: 1. [unit]: [DRIFT: type] -- [brief description] 2. [unit]: [DRIFT: type] -- [brief description] 3. [unit]: [DRIFT: type] -- [brief description] [If --report]: Full report saved to .manuscript/translation/{lang}/BACK-TRANSLATION-REPORT.md [If high drift]: Consider re-translating problem units or reviewing with /scr:cultural-adaptation. ``` ## Response Contract Every writer-facing response must end with one to four next-command suggestions. Each suggestion must include a short explanation of what that path will do. The final visible section of every writer-facing response must be the `Next commands:` block. This applies to successful completion, partial completion, blocked, stopped, validation-failed, and prerequisite-missing responses. Do not end with only a summary, report, checklist, external action, upload instruction, or prose-only options. Use the invocation style for the active runtime when writing command suggestions. Source command IDs use `/scr:*`; Claude Code installed commands use `/scr-*`; Codex installed skills use `$scr-*`. Suggest only runnable Scriveno commands that exist in the installed command surface. Do not invent adjacent workflow names. Use this format: ```markdown Next commands: - `/scr:...`: One short sentence explaining what this path will do. - `/scr:...`: One short sentence explaining what this alternate path will do. ``` If exactly one path is clearly best, provide one suggestion. If two, three, or four useful paths exist, show them as alternatives. Do not force a linear path when the writer has a real choice. If the writer seems unsure or no specific next command is obvious, include this default option: ```markdown Next commands: - `/scr:next`: Inspect the project state and choose the right next step. ``` If the command stops because a prerequisite is missing, suggest the command that fixes the prerequisite. Keep every explanation practical and writer-facing.