Natural Language Understanding

  • Release version: Xanadu
  • Updated August 1, 2024
  • 4 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 Natural Language Understanding

    ServiceNow® Natural Language Understanding (NLU) enables your system to learn and respond to human-expressed intent by processing natural language inputs. Through the NLU Workbench and inference service, you can create, train, and deploy models that interpret user or system actions based on utterances—natural language examples representing user intents. This allows your applications to infer meaning and context from text such as chat entries, incident descriptions, or emails.

    Show full answer Show less

    Key Features

    • NLU Workbench: A platform for building, training, testing, and publishing NLU models. It supports creation of intents and entities from natural language utterances and allows iterative improvement of models.
    • Inference Service: Translates user utterances into machine-understandable formats by predicting intents and entities, enabling intelligent system actions triggered by natural language.
    • Model Management: Supports the full lifecycle of NLU models including multilingual support, allowing consistent understanding across multiple languages.
    • Integration with Virtual Agent: Virtual Agent administrators can configure conversation flows that consume NLU models for enhanced chatbot understanding, and update models directly within the Virtual Agent Designer interface.
    • Advanced Features: Extended functionalities in the NLU Workbench help manage and improve model performance.

    Key Components and Terminology

    • Intent: The action or goal a user wants to perform (e.g., granting access).
    • Utterance: Natural language examples used to train intents; should be clear and unambiguous.
    • Entity: Contextual objects related to intents, such as devices, roles, or priority levels.
    • System Entities: Predefined entities like date, time, and location for reuse across models.
    • User Defined Entities: Custom entities created from utterance words to tailor models to your domain.
    • Vocabulary: Mechanism to define or override word meanings, including synonyms.
    • NLU Model: A collection of utterances with associated intents and entities used for inference.

    Practical Benefits for ServiceNow Customers

    By leveraging NLU, you can enhance your ServiceNow applications with natural language capabilities, improving user experience and automating intent recognition. This enables better interaction with virtual agents, more accurate processing of user requests, and supports multilingual environments for global deployments. The integration and management tools provided allow for seamless model updates and continuous improvement without requiring extensive technical expertise.

    ServiceNow® Natural Language Understanding (NLU) provides an NLU Workbench and an NLU inference service that you can use to enable the system to learn and respond to human-expressed intent. By entering natural language examples into the system, you help it understand word meanings and contexts so it can infer user or system actions.

    Overview of NLU

    Figure 1. User input flow in the NLU model build process
    This image shows you the user input flow in the NLU model build process.
    This image shows you the user input flow in the NLU model build process.

    NLU terminology

    In NLU parlance, these terms identify the key language components the system uses to classify, parse, and otherwise process natural language content.
    Intent
    Something a user wants to do or what you want your application to handle, such as granting access.
    Utterance
    A natural language example of a user intent. For example, a text string in an incident's short description, a chat entry, or an email subject line. Utterances are used to build and train intents and should therefore not include several or ambiguous meanings or intents.
    Entity
    The object of, or context for, an action. For example: a laptop, a user role, or a priority level.
    System entity
    These are predefined in an instance and have highly reusable meanings, such as date, time, and location.
    User defined entity
    These are created in the system by users and can be built from words in the utterances they create.
    Common Entity
    A context commonly used and extracted via a pre-defined entity model, such as currency, organization, people, or quantity.
    Vocabulary
    Vocabulary is used to define or overwrite word meanings. For example, you can assign the synonym “Microsoft” to the acronym “MS”.
    NLU Model
    A collection of utterance examples and their associated intents and entities that the system uses as a reference to infer intents and entities in a new utterance. The NLU Workbench comes with pre-built NLU models for specific business units, such as an ITSM model. You can also create custom models.

    This image illustrates how Natural Language Understanding processes and renders utterance examples into intents and entities in the system.

    Figure 2. NLU processes and renders utterance examples into intents and entities
    This image shows you how Natural Language Understanding processes and renders utterance examples into intents and entities in the system.

    NLU Workbench

    Use the NLU Workbench to create morphological representations of human language. These models enable you to create intents and entities expressed in natural language utterances. Any ServiceNow application can invoke an NLU model to get an inference of intents and entities in a given utterance.

    Using the nlu_admin role, you build your models in the NLU Workbench, where you create, train, test, and publish them iteratively.

    Figure 3. Overview of NLU Authoring API helping administrators build their models
    This image shows you how the NLU Authoring API helps NLU administrators build their NLU models.

    For information on how to build and use an NLU model, see: Create an NLU model.

    NLU inference service

    Natural Language Understanding provides an NLU inference service that 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 model so that its text translates into machine-understandable formats, such as APIs and parameters.

    Figure 4. Overview of how the system uses an NLU inference API to extract intents and entities
    This image shows you how the system uses an NLU inference API to extract intents and entities for a given utterance. This image shows you how the system uses an NLU inference API to extract intents and entities for a given utterance.

    Here, the system uses an inference API to train NLU algorithms by using sample record data to identify intents and entities that are strong candidates for accurate prediction.

    NLU model consumption

    Other ServiceNow® applications consume NLU model output, such as Virtual Agent.

    Figure 5. Overview of Virtual Agent application consuming NLU
    This image shows you how the Virtual Agent application consumes Natural Language Understanding.

    For example, Virtual Agent administrators can configure a Virtual Agent Designer conversation flow to consume NLU models so that agent chatbots can better understand user statements in the conversation. For more information on how Virtual Agent consumes NLU models, see: Natural Language Understanding (NLU) topic discovery in Virtual Agent.

    Get started

    Troubleshoot and get help