Translating with Google AutoML
If you use Google's AutoML MT, you can integrate it into Smartling and use it as an MT for a Translation workflow step and even as the MT resource in the CAT Tool.
Using Smartling's MT Profile feature, you can securely store the Google AutoML MT credentials in Smartling, and assign the profile to any MT workflow step and even as the MT resource in the CAT Tool. The required credentials are found in an exported JSON file from your Auto ML account. For steps on how to export this JSON file, read the AutoML documentation.
To successfully integrate with SMartling, the following permissions should be granted:
Creating an MT Profile in Smartling
- In the Account Settings dropdown, click on Integrations
- Alternatively, you can access the MT Profiles page from the Project Settings > Workflow Configuration > Translation step > Manage Step > Create MT Profile
- Click on the MT Profiles tab
- Click on Create MT Profile
- Define a name under Profile Name
- The name is visible in the Smartling UI
- Choose Google AutoML Translation from the MT Provider dropdown
- Complete all required fields by entering the credentials from the exported JSON file as follows;
- Location: your location will either be
JSON credentials: copy and paste the entire JSON file into this field, including the brackets
Model ID: This is the custom model name for the entire MT engine, if applicable.
- Custom Language Code: Use this fields only if your custom MT model uses locales which are not supported by Google AutoML, insert the custom language code here.
- Leave blank to use default language code mapping
- Glossary ID: is custom glossary name, if applicable
- Both the Smartling Glossary and your AutoML Glossary work together, meaning the two lists of terms are referenced and used in your translations
- Insert the entire path, e.g.:
- (Optional) Click Test Credentials to ensure the credentials are correct.
- Ensure all fields are complete before testing (including the Profile Name).
- (Optional) Define language-specific credentials and rules.
- This only applies if you have multiple custom models/glossaries and want to use them only for specific language pairs but not others.
- Bilateral Mode: enabling this feature allows for specific language pairs to be machine translated as both source > target and target > source.
- This allows you to reverse the language pairs and leverage source content as translations