What This Agent Does
This Document Translation Quality Assurance workflow automates the process of translating Japanese documents to English while maintaining rigorous quality standards through back-translation verification. The agent reads Japanese files from Google Drive, translates them to English using Claude AI, back-translates the English version to Japanese, and generates a comprehensive quality report comparing the original and back-translated versions to identify translation discrepancies.
Key benefits and time savings:
-
Eliminates manual quality checks by automating the back-translation comparison process
-
Saves 2-3 hours per document by streamlining translation verification workflows
-
Ensures translation accuracy through intelligent diff analysis that highlights potential meaning shifts
-
Creates audit trails with detailed reports stored in Google Docs for compliance and review purposes
-
Enables rapid feedback by notifying teams instantly via Slack when quality reports are ready
Target use cases:
- Legal and compliance document translation requiring high accuracy
- Technical documentation localization for Japanese software products
- Multilingual content management for international businesses
- Quality assurance in translation services and localization agencies
- Academic and research paper translation verification
Who Is It For
This workflow is ideal for:
-
Translation teams and agencies managing high-volume Japanese-to-English projects
-
Multinational corporations requiring quality-assured document translation
-
Legal and compliance departments needing verified translations for official documents
-
Software localization teams ensuring accurate technical documentation
-
Content managers maintaining translation consistency across multiple documents
-
Quality assurance specialists implementing systematic translation verification processes
Required Integrations
Google Drive
Why it's needed: Google Drive serves as your document repository, allowing the workflow to automatically access and read Japanese source documents without manual file transfers.
Setup steps:
- Ensure you have a Google account with Google Drive access
- Create a dedicated folder in Google Drive for Japanese source documents (e.g.,
/Japanese Documents/)
- Upload your Japanese documents to this folder
- In TaskAGI, navigate to Integrations → Google Drive
- Click Connect and authorize TaskAGI to access your Google Drive
- Grant permissions for reading and creating files
- Verify the connection by testing file access
How to obtain credentials:
- Google Drive uses OAuth 2.0 authentication
- TaskAGI handles the OAuth flow automatically when you click Connect
- No manual API key entry is required
- Your authorization is securely stored in TaskAGI
Configuration in TaskAGI:
- Select your Google Drive account from the dropdown
- Specify the folder path containing Japanese documents
- Ensure the service account has read permissions on the target folder
Anthropic (Claude AI)
Why it's needed: Anthropic's Claude API powers the intelligent translation and back-translation processes, ensuring high-quality, context-aware language conversion.
Setup steps:
- Visit console.anthropic.com
- Create an Anthropic account or log in to your existing account
- Navigate to API Keys in the left sidebar
- Click Create Key to generate a new API key
- Copy the generated key (you'll only see it once)
- In TaskAGI, go to Integrations → Anthropic
- Click Add Integration and paste your API key
- Name the integration (e.g., "Claude Translation Service")
- Click Save and Test to verify the connection
How to obtain API keys:
- Visit Anthropic's console
- Set up billing information (Claude API is pay-as-you-go)
- Generate API keys from the API Keys section
- Keep your key confidential and never share it publicly
Configuration in TaskAGI:
- Model selection:
claude-sonnet-4-5-20250929 (pre-configured for optimal translation quality)
- Temperature: Set to
0.3 for consistent, accurate translations
- Max tokens: Set to
4000 to accommodate longer documents
Google Docs
Why it's needed: Google Docs provides a professional, shareable platform for generating and storing quality assurance reports with formatting, comments, and version history.
Setup steps:
- Use the same Google account connected to Google Drive
- In TaskAGI, navigate to Integrations → Google Docs
- Click Connect and authorize TaskAGI to create and edit documents
- Grant permissions for document creation and modification
- Create a dedicated Google Drive folder for reports (e.g.,
/Translation QA Reports/)
- Note the folder ID from the URL (appears after
/folders/)
- Configure the folder ID in TaskAGI's Google Docs settings
How to obtain credentials:
- Google Docs uses the same OAuth 2.0 authentication as Google Drive
- No separate API key is needed
- TaskAGI manages authentication automatically
Configuration in TaskAGI:
- Set the default folder for report creation
- Configure document naming convention:
Translation QA Report - [Date] - [Document Name]
- Enable sharing settings to make reports accessible to team members
Slack
Why it's needed: Slack integration enables real-time notifications to your team when translation quality reports are completed, keeping everyone informed and enabling rapid review cycles.
Setup steps:
- Open your Slack workspace (you must be an admin or have app installation permissions)
- Visit api.slack.com/apps
- Click Create New App → From scratch
- Name your app (e.g., "TaskAGI Translation Agent")
- Select your workspace
- In the left sidebar, click OAuth & Permissions
- Under Scopes, add these bot token scopes:
-
chat:write
-
files:read
-
files:write
- Click Install to Workspace and authorize the app
- Copy the Bot User OAuth Token (starts with
xoxb-)
- In TaskAGI, go to Integrations → Slack
- Paste your Bot User OAuth Token
- Click Save and Test
How to obtain API keys:
- Slack uses OAuth 2.0 with bot tokens
- The Bot User OAuth Token is generated automatically after app installation
- Store this token securely in TaskAGI
Configuration in TaskAGI:
- Select your Slack workspace
- Choose the notification channel (e.g.,
#translation-qa)
- Configure message format to include document name, status, and report link
- Enable thread replies for organized discussions
Configuration Steps
Node-by-Node Configuration Guidance
1. Input Form (Trigger)
This node captures the document selection from users.
-
Field name:
japanese_document_url
-
Field type: Text input or file picker
-
Description: "Paste the Google Drive link to your Japanese document"
-
Example value:
https://drive.google.com/file/d/1ABC123XYZ/view
2. Read Japanese Document
Retrieves the content from the specified Google Drive file.
-
File ID source: Extract from the input form URL
-
Output variable:
original_japanese_text
-
Expected output: Full text content of the Japanese document
3. Translate JP to EN
Converts Japanese text to English using Claude.
-
Input:
original_japanese_text
-
Prompt template: "Translate the following Japanese text to English, preserving all technical terms and formatting: [TEXT]"
-
Model:
claude-sonnet-4-5-20250929
-
Output variable:
english_translation
-
Example: Japanese legal document → English legal document with maintained terminology
4. Back-Translate EN to JP
Translates the English version back to Japanese for comparison.
-
Input:
english_translation
-
Prompt template: "Translate the following English text back to Japanese, maintaining the same style and terminology: [TEXT]"
-
Model:
claude-sonnet-4-5-20250929
-
Output variable:
back_translated_japanese
5. Generate Diff Analysis
Compares the original Japanese with the back-translated Japanese to identify discrepancies.
-
Input 1:
original_japanese_text
-
Input 2:
back_translated_japanese
-
Function type: String comparison with diff highlighting
-
Output variable:
quality_analysis
-
Metrics generated:
- Similarity percentage
- Key differences identified
- Risk assessment (high/medium/low)
6. Create Quality Report
Generates a new Google Docs document for the report.
-
Document title:
Translation QA Report - [Timestamp] - [Document Name]
-
Folder:
/Translation QA Reports/
-
Output variable:
report_document_id
7. Add Report Content
Populates the document with comprehensive analysis.
-
Content sections:
- Executive Summary
- Original Japanese text (first 500 characters)
- English translation
- Back-translated Japanese
- Diff analysis with highlighted discrepancies
- Quality score and recommendations
- Timestamp and metadata
8. Send Slack Notification
Alerts the team with report completion details.
Testing Your Agent
Step 1: Prepare Test Data
- Upload a short Japanese document (200-300 words) to Google Drive
- Choose a document with mixed content (regular text + technical terms)
- Note the file URL for use in testing
Step 2: Execute Test Run
- In TaskAGI, open your workflow
- Click Test Execution
- Paste your test document URL in the input form
- Click Run Workflow
- Monitor the execution progress in real-time
Step 3: Verify Each Step
-
Read Japanese Document: Confirm text extraction is complete and accurate
-
Translation: Review English output for grammar and meaning preservation
-
Back-Translation: Check if the Japanese version closely matches the original
-
Diff Analysis: Verify that discrepancies are correctly identified
-
Report Creation: Confirm Google Docs document is created in the correct folder
-
Slack Notification: Check that your team receives the notification with the correct link
Step 4: Expected Results and Success Indicators
✅ Successful execution shows:
- All 9 processing nodes complete without errors
- Quality score above 85% for well-translated documents
- Report document created and accessible in Google Drive
- Slack notification delivered within 30 seconds
- No authentication or permission errors
- Complete data flow from trigger through all processing nodes to output
Troubleshooting tips:
- If Google Drive read fails, verify file sharing permissions
- If translation is slow, check your Anthropic API quota and billing status
- If Slack notification doesn't arrive, confirm the bot token has
chat:write scope
- If diff analysis shows unexpected results, review the original document encoding
Your workflow is now ready for production use! Start with small batches of documents to validate quality standards before scaling to larger volumes.