You may have already activated pre-built Virtual Agent topics or created your own topics. After success with one initial language, you may decide it’s time to go international!
Previously, localizing topics was a manual process of translating individual strings. In the Rome release, this is no longer the case. You can enable dynamic translation to automatically translate entire Virtual Agent topics. You can also automatically translate NLU models and all the utterances within. This lets you deliver a more personal conversational experience to your global workforce without having to configure individual languages.
To automatically translate topics and NLU models, install the Dynamic translation plugin and connect with a 3rd party translation service (IBM, Microsoft, or Google). Virtual Agent topics can be translated into any language available on the ServiceNow platform, which is over 20 languages. NLU intents and utterances are supported in 16 languages, with entities supported in 5 languages (English, French, Spanish, German, and Japanese).
If you’re using pre-built ITSM Virtual Agent conversations and NLU models from the ServiceNow Store, then French and German are automatically available. More languages are coming in Winter 2021.
Required plugins to translate your Virtual Agent topics and NLU models:
- Virtual Agent, NLU (part of Pro or Enterprise SKU)
- Dynamic Translation (part of Pro or Enterprise SKU)
- Rome+ instance
- Recommended: ITSM Virtual Agent Conversations and NLU models ServiceNow store apps for a head start
- i18n language pack(s)
Sign up for a 3rd party translation service
Sign up and create an account with a cloud-based translation service provider. By default, the following translation service providers are supported. Note that these services are billed at a per usage rate, however their free tiers are very generous:
- IBM: To sign up, see the IBM website.
- Microsoft: To sign up, see the Microsoft documentation.
- Google: To sign up, see the Google website and Set up Google Cloud Translator Service spoke.
Activating dynamic translation
To enable dynamic translation of Virtual Agent topics and NLU models, install the plugins and sign up for one of the services noted above. Then follow one of the following docs depending on the translation service you selected:
- Activate dynamic translation with IBM
- Activate dynamic translation with Microsoft
- Activate dynamic translation with Google
Translating your Virtual Agent topics
Navigate to the Virtual Agent Designer, and navigate to a topic’s Languages and Translations tab. If you installed an i18n language pack, you’ll see the language listed on the Translations tab. Otherwise, you need to manually create a language record in the sys_language table. Click “Edit” next to the desired language, and then click the “Machine Translate” button. Your topic will then be translated in the selected language. You can then review and publish your translated topic.
Quick tip!: You can now just as easily test your topic in different languages without switching languages in the user profile. In your tester’s URL, just add “&sysparm_requester_session_language=<language code, e.g., ‘es’ for Spanish>”
Translating your NLU models
To translate NLU models, navigate to the NLU Workbench. On the right-hand side of a model name, click the 3-dot menu and select “Add a language”. One of the options is “Use software”. This automatically translates the model to the language selected in the drop-down.
NLU models in your translated language will appear as a “secondary” model. Within each intent, you can review both the translated and original utterance. You can review and publish your model.
A final disclaimer: you should still use a set of eyes to review dynamic translations. Translations from any machine/model may not account for unique phrasing and the idiosyncrasies of language. A topic author or NLU model author should take the opportunity to review and refine any translated content if necessary. If you prefer a hands-on approach, you can also rely on the Localization Framework which provides a workflow tool to assign Virtual Agent topics and NLU models to be manually translated by a linguist. More information on using the Localization Framework for translations can be found here.
Here's a Virtual Agent Academy session with a step-by-step walkthrough:
Try it out in your Rome instance today!
