Localization options for Virtual Agent
Summarize
Summary of Localization options for Virtual Agent
The ServiceNow platform offers multiple methods to support international Virtual Agent users by localizing conversations in various languages and locales. You can choose between professional localization by expert linguists, dynamic machine translation, or a hybrid approach combining both. These options enable you to tailor the user experience based on your audience’s language diversity, balancing translation quality, implementation cost, and scalability.
Show less
Direct Language Support with Professional Localization
This approach involves expert linguists translating and configuring entities, intents, strings, and vocabulary directly within the Assistant Designer Asset library. It provides the highest translation accuracy and allows defining custom vocabulary to improve natural language understanding (NLU).
- Benefits: More accurate translations, localized NLU intents and entities for supported languages, better user experience.
- Constraints: Higher cost, longer development time, requires ongoing maintenance, limited scalability and extensibility.
- Currently, 5 languages have full NLU intent and entity support, and 12 languages have NLU intent support only.
Dynamic Language Detection and Machine Translation
Dynamic translation uses third-party services to translate user utterances and Virtual Agent responses in real time, enabling faster deployment and lower costs. However, translation accuracy depends on the service and may vary, especially where contextual clues are limited.
- Benefits: Lower implementation costs, quicker setup, less human maintenance, support for many languages, scalable and extensible.
- Constraints: Limited NLU support (5 languages with full intent and entity support, 12 with intent only), possible reduced user experience quality, potential mistranslations.
- Supports 22 language plugins with basic translations; additional unsupported languages can be dynamically translated but with variable quality.
Multi-language Deployment Strategies
Organizations can combine professional localization for high-priority languages with dynamic translation for others to optimize both user experience and coverage. You can also designate fallback languages for locale variants (e.g., Mexican Spanish falling back to Spanish) to enhance topic, keyword, and NLU prediction accuracy.
Localization Insights and Management
ServiceNow provides a Localization Insights dashboard accessible via the Assistant Designer Asset library to monitor translation and localization metrics. The Localization Framework helps manage the translation process whether using professional, machine, or hybrid approaches.
Enabling Dynamic Translation in Virtual Agent
To use dynamic machine translation, you must:
- Install the relevant ServiceNow® language and Dynamic Translation plugins.
- Configure translation service credentials and set the default translation and language detection provider.
- Enable Dynamic Translation for selected languages in Virtual Agent Chat Settings.
- Activate multilingual service support in Now Assist applications for native or dynamic translations.
- Specify fallback languages for locale-specific language support to improve NLU topic matching and user experience.
Practical Impact for ServiceNow Customers
By leveraging these localization options, ServiceNow customers can effectively support a global user base with Virtual Agent conversations tailored to multiple languages and locales. Selecting the right mix of professional and dynamic translation methods allows balancing translation quality, deployment speed, scalability, and cost. Proper configuration and use of fallback languages enhance NLU accuracy and conversational relevance, ensuring better engagement and user satisfaction across diverse linguistic audiences.
The ServiceNow platform provides several methods for serving your international Virtual Agent users, regardless of their language and locale. Professionally localize your conversations, use dynamic machine translation, or a combination of the two.
Direct and dynamic language support
If your chat audience is diverse, you want to support them in their own language whenever possible. One option is to localize everything professionally using linguists who can translate your topics into the desired languages. While this option produces the best user outcomes, it is often the most expensive. Another option available to you is language detection combined with dynamic machine translation. Instead of using translated topics and models, dynamic translation connects to a third-party machine translator. During the conversation, user utterances and responses are translated into the primary language on the instance> All results are translated back into the user's language for the session. There are benefits and drawbacks for all options, as described in the following list. Before activating machine translations, you should carefully weigh the costs and benefits and set proper expectations for your deployment objectives.
- Direct language support (professional localization)
- A direct language strategy employs an expert linguist to configure entities, intents,
strings, keywords, and vocabulary, which are then used to match user utterances with
configured content. Since an expert provides translations that directly match the
conversation's content, the accuracy of the translated content will be high. In addition to
the translations themselves, experts can use the
NLU vocabulary feature
to further define words and phrases that
it may encounter from your users. For example, a user may enter an acronym such as
MS, which you can define as the vocabulary item
Microsoft.
Professional translations are implemented directly in Assistant Designer Asset library. For details, see Localizing Virtual Agent conversations.
Table 1. Using translations from professional linguists and corresponding ServiceNow language plugins Benefits Constraints - More accurate translations
- For supported languages, NLU intents and entities are localized
- Requires more resources to build and maintain
- Longer completion time
- Higher implementation costs
- 5 languages have NLU intent and entity support
- 12 languages have NLU intent support only
- Increased coverage is not typically scalable
- Not easily extensible
For more information, see Language support for NLU services.
- Dynamic language detection and machine translation
- Dynamic machine translation relies on the translation service to translate all content in
real time. The accuracy of the translations is dependent on the translation service. Although
translation services typically leverage algorithms to translate words based on the context of
the conversation or sentence, there may be times when there is minimal context for the
algorithm to leverage. Accuracy may be diminished in this situation. Check with your
translation service provider about the ability to exclude words from translation or associate
vocabulary with specific synonyms.
Table 2. Using dynamic translation with the supported ServiceNow language plugins Benefits Constraints - Lower implementation costs
- Faster completion time
- Less human maintenance
- ServiceNow provides 22 language plugins with basic translations
- Limited scalability
- Limited extensibility
- 5 languages have NLU intent and entity support
- 12 languages have NLU intent support only
For more information, see Language support for NLU services.
Dynamic machine translation for languages that are not supported on the ServiceNow platform provides coverage, but the quality of the user experience may be lower.Table 3. Using dynamic translation with unsupported languages Benefits Constraints - Lower implementation costs
- Faster completion time
- Less human maintenance
- Support for many languages
- Highly scalable
- Highly extensible
- User experience may vary
- Potential for mistranslation may reduce user outcomes
Multi-language deployment strategies
Depending on the needs of your user base, you may want a mix of solutions to maximize both user experience and access. For example, you may choose to professionally localize prioritized languages and dynamically translate others. For example, if 95% of your users speak six languages, you may decide to use a linguist to translate your topics into those six. You might then use Dynamic Translation for the remaining five percent.
You can also enable locale support by designating translated languages as fallback languages. For example, you can support Mexican Spanish users by designating Spanish as the fallback language for that locale.
In the following example, French, German, and Spanish are professionally localized, including the NLU models, intents, and entities. This localized content provides the best user experience. Polish is supported as a language on the ServiceNow platform, and when the plugin is installed and activated, it includes some translated content. In this imaginary scenario, the business decision was made to dynamically translate Polish content because it accounts for a smaller percentage of users. The quality of this experience may vary since NLU models are not localized. Instead, utterances are dynamically translated to English, and NLU topic discovery is run through the English model. Since Czech is currently not a supported language on the ServiceNow platform, dynamic machine translation is the only option. Admins can add Czech as a language and enable Dynamic Translation for the language in Virtual Agent.
Localization insights
You can view dynamic translation and localization insights from the Assistant Designer Asset library. Navigate to , select the Asset library tab, and select Localization Insights from the list of links in the side panel.
For more information, see Localization Insights dashboard.