NLU Service updates

  • Release version: Yokohama
  • Updated January 30, 2025
  • 2 minutes to read
    • Summarize
    Summarized using AI
    This content was generated using new OpenAI-powered functionality. Results are provided on an as is basis and are not guaranteed to be accurate or complete.

    Summary of NLU Service updates

    The NLU Service in ServiceNow enhances your system's ability to understand natural language by training and predicting intents and entities from user utterances. It supports both spoken and written language inputs to drive intelligent actions, particularly within Virtual Agent conversations. This service is updated independently from your instance upgrades on a bi-monthly basis and includes both minor automatic updates and major updates aligned with family releases (e.g., Rome, San Diego, Tokyo).

    Show full answer Show less

    These updates improve the quality and accuracy of NLU model training and predictions without requiring any action from you. While most updates do not affect existing NLU usage, some changes may impact how you manage or configure your models.

    Key Features

    • Dialog Acts (May 2023): Introduced support for dialog acts such as affirm, negate, and modify to enable more natural mid-conversation interactions in Virtual Agent. Enabled by default for English in new VA topics.
    • Language Model Enhancements: Migrated all languages to new models, improving average intent prediction quality by 10%. Upgraded Scandinavian languages’ models (Danish, Swedish, Finnish, Norwegian) with a 17% quality improvement as of January 2023.
    • Irrelevance Management: Allowed customers to edit and manage irrelevant utterances to enhance irrelevance detection in models.
    • Model Flexibility: Removed the requirement for models to have multiple intents, simplifying end-to-end testing of single-intent Virtual Agent topics.
    • System Entity Improvements: Added a new DATE-TIME system entity (English only) and expanded system entity support to Portuguese and Brazilian Portuguese.
    • Vocabulary and Entity Support Expansion: Extended vocabulary source and entity support for Dutch and Italian languages, including simple, mapped, and open-ended entities.
    • Performance Enhancements: Improved intent and entity detection by better handling common words, punctuation, and special characters. Optimized latency and memory usage for Named Entity Recognition (NER) system entities.
    • Version Support Policy: Customers must use newer NLU Service versions and cannot reference versions older than two releases behind their current Glide version.
    • Model Training Improvements: Incorporated admin feedback from NLU Workbench to refine intent prediction and training data quality.

    What This Means for ServiceNow Customers

    • You benefit from continuous improvements in NLU accuracy and performance without needing manual updates, ensuring your Virtual Agent and other NLU-driven processes remain effective.
    • New features like dialog acts and expanded language support allow for richer, more natural conversations and broader multilingual capabilities.
    • Greater control over irrelevant utterances and simplified model requirements help streamline model management and testing efforts.
    • Awareness of version support policies ensures your platform remains compatible with the latest NLU Service capabilities.

    Refer to this documentation so you are up to date with changes to the NLU Service.

    Service update summary

    The NLU Service helps the system to understand natural language and drive intelligent actions. This service trains and predicts intents and entities for a given user utterance in your NLU model so it can understand human-expressed natural language, whether spoken or written. The source of this documentation is KB0953693.

    This service is updated independently of your instance upgrade, and without any action required by you. These updates are done on a bi-monthly basis (once every two months) to improve the quality of NLU model training and predictions. Major updates are aligned with family releases such as Rome, San Diego, Tokyo, etc. Minor updates are automatically updated so you are using the latest version when you retrain an NLU model. While most of these updates don't impact your existing use of NLU, there may be some changes you need to be aware of.

    May 2023 NLU Service update

    • Introduced dialog acts to enable natural mid-conversation in Virtual Agent (VA) and improve conversation fluidity. Affirm, negate, and modify dialog acts are supported in English and enabled by default for all new VA topics.
    • Migrated all languages to use new language models, boosting average intent prediction quality by 10% across all languages.
    • Enabled customers to manage and edit irrelevant utterances for their models to improve irrelevance detection.
    • Removed the requirement for a model to have 2 or more intents in the model, making it easier for end-to-end topic testing in VA.

    March 2023 NLU Service update

    • Improved intent/entity detection through better handling of common words in vocabulary sources.
    • Improved latency and memory utilization for system entity (NER) detection.
    • Updated version support so that customers will need to use newer versions of the NLU Service and cannot point to n-2 releases older than their current glide version.

    January 2023 NLU Service update

    • Created the new DATE-TIME system entity (English only) for use in Virtual Agent.
    • Added vocabulary source and entity support (simple, mapped, and open ended) for the Dutch and Italian languages, and system entity support for Portuguese and Brazilian Portuguese.
    • Upgraded the ServiceNow Language Model for the Danish, Swedish, Finnish, and Norwegian languages, improving their average prediction quality by 17% from Tokyo.
    • Improved the handling of punctuation in entities and special characters in vocabulary sources.
    • Incorporated feedback provided on intent predictions by admin users in the NLU Workbench to improve model training data for Virtual Agent.