NLU models

  • Release version: Xanadu
  • Updated August 1, 2024
  • 3 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 models

    NLU models enable ServiceNow customers to apply Natural Language Understanding (NLU) on their instances, allowing the system to interpret and respond to user natural language inputs. Each model consists of utterances (user input examples), intents (system actions triggered), entities, and vocabulary used to train the model to map user requests to appropriate system actions.

    Show full answer Show less

    Users with the nluadmin role can create, manage, test, and publish these models via the NLU Workbench interface, which organizes models by application type: Virtual Agent, Issue Auto Resolution (IAR), and AI Search.

    Key Features

    • Multi-application support: NLU models can be used across Virtual Agent, Issue Auto Resolution, and AI Search applications.
    • Model management: The NLU Workbench allows filtering models by language, viewing multilingual model groups, and identifying models linked to specific applications.
    • Prebuilt models: Read-only prebuilt models are available to copy and customize, especially for Virtual Agent and AI Search; IAR uses a prebuilt model configured through its Admin console.
    • Model tuning and performance: The Workbench provides sections and cards to access performance-boosting functions tailored per application type.
    • Role-based access: Only users with the nluadmin role can access these model management features.

    Application-Specific Details

    • Virtual Agent: Displays a list of models with options to view or copy prebuilt models. Includes five performance enhancement functions accessible via the Workbench.
    • Issue Auto Resolution (IAR): Uses a single prebuilt model configured externally. Management is accessed via IAR Tuning; model creation and performance boosting options are not available in the Workbench.
    • AI Search: Provides one prebuilt model that can be copied and customized. The Workbench shows two performance function cards and allows new model creation.

    Next Steps

    To effectively manage and optimize NLU models, customers should explore the model life-cycle and content management documentation. This will help in creating tailored models that enhance natural language interactions across ServiceNow applications.

    Use NLU models to apply ServiceNow Natural Language Understanding on your instances. Create, manage, test, and publish NLU models with the NLU Workbench.

    Figure 1. Overview of how user input is transformed to a system action
    Diagram for how user input is transformed into a system action.

    A model is a collection of utterances, intents, entities, and vocabulary that the system uses to respond to natural language inputs from your users. The model takes the natural language input from your users and matches it to actions to be performed by the system.

    The utterances in your model data are examples of what your users might ask for. These examples are used to train your model to recognize which system actions to perform in response. These system actions are called intents.

    Usage, roles, and navigation

    Use the nlu_admin role to access the NLU Workbench.

    Navigate to NLU Workbench > Models.

    There are three tabs on the NLU Workbench homepage: Virtual Agent, Issue Auto Resolution, and AI Search. The Virtual Agent tab opens by default. Select the correct tab for your application.

    The homepage view of the Natural Language Understanding Workbench.

    You can use the Search field to look for models by name.

    The default setting of the All Languages filter displays all models regardless of their language. You can filter models by language using the list in All Languages. To learn more about the available languages, see NLU language support.

    If you have a multilingual grouped model, its row in the list of models has an arrow on the left side. Select the arrow to expand the group so that all language versions of the model are displayed. For more information on model languages and grouping, see Multilingual model management.A grouped model expanded to show two language models, one for English and one for Brazilian Portuguese.

    The Used In VA column indicates whether a model has already been linked to another application such as Virtual Agent. For AI Search, this column indicates whether the model has already been linked to a Genius Result. Similarly, the Mapped Intents column displays a count of intents that have been mapped or linked to Virtual Agent topics.

    Scroll down the NLU Workbench homepage to view the Boost your model performance section. This section displays cards you can use to access functions that are available to your model type.The Boost your model performance section of the model homepage, with two cards displayed.

    Model applications

    The output of your NLU models can be consumed by the ServiceNow applications Virtual Agent, Issue Auto Resolution, and AI Search. The NLU Workbench homepage organizes each application's models by tab. In the list of models, different columns are displayed depending on the application.

    NLU Models for Virtual Agent

    For more information on Virtual Agent, see Virtual Agent.

    • The tab for Virtual Agent displays a list of models. Select a model name to open an overview page for that model.
    • The Show Prebuilt Models toggle is off by default. Switch this toggle on to display read-only models that can be copied and used as the basis for your custom models.
    • The Boost your model performance section displays a total of 5 cards for functions available to Virtual Agent.

    NLU Models for IAR

    Fewer functions and columns are displayed for IAR than for the other applications in NLU Workbench. This is because the IAR model is prebuilt and is configured in the IAR Admin console.
    • When you select the name of your IAR model in NLU Workbench, you are taken to IAR Tuning rather than to a model overview. For more information, see Issue Auto Resolution Tuning in NLU.
    • The Show Prebuilt Models toggle is not displayed.
    • The Create new model button is not displayed because the prebuilt model is used directly.
    • The section Boost your model performance is not displayed.

    For more information on setting up IAR, see Using Issue Auto Resolution.

    NLU Models for AI Search

    • There is one prebuilt model for AI Search. You can copy this model as the basis for a custom model.
    • The Show Prebuilt Models toggle is on by default.
    • The Create new model button is available.
    • The Boost your model performance section displays two cards for functions available to AI Search.

    For an introduction to managing AI Search in the NLU Workbench, watch the following video.

    For more information, see the AI Search landing page.

    What you can do next

    To learn more about model content and the model life-cycle, see Model management.