NLU Models | Natural Language Understanding
By: Ethan B. 

Overview and Activation

Natural Language Understanding (NLU) is a feature in ServiceNow in which a system such as a Virtual Agent bot is trained to understand unstructured user input. The system uses an NLU model to intelligently predict what a user is trying to communicate. Using NLU can greatly improve users’ Virtual Agent experience.

To use Natural Language Understanding, an administrator must activate the following plugins:

• NLU Workbench – Core [com.glide.nlu]
• NLU Workbench [com.snc.nlu_studio]
• Predictive Intelligence [com.glide.platform_ml]

How It Works

An NLU model contains a list of intents, which are encapsulations of what a user might want the Virtual Agent to help them accomplish. Each intent is associated with a list of utterances, which are user inputs that the model is trained to understand. The model compares the user’s input against its list of utterances to determine the user’s most likely intent. An NLU model also contains a list of entities that it is trained to recognize within a user’s utterance. Entities are the important details of a request that the system needs to know in order to carry out the task that the user desires.

For example, a user tells the VA that they downloaded an application from the company, but it isn’t working. “My application isn’t working” would be the utterance, and the name of the application would be the entity parsed from the utterance. The associated intent that the NLU model recognizes might be to submit a support ticket on their behalf, using the entity (the application name) as a detail to include in the ticket.

Intents can be mapped to Virtual Agent topics, allowing the VA to provide different dialogue flows depending on a user’s intention. Having multiple utterances associated with each intent increases the chance that the bot will understand the user’s intention, regardless of variations in a user’s input.

NLU Workbench

ServiceNow’s interface for creating, viewing, and modifying NLU models is the NLU Workbench. To access the NLU Workbench, navigate to NLU Workbench > Models through the All menu. Within the NLU Workbench, a developer can add new intents, associate utterances with them, configure entities to capture important context, and train and test the model before deployment.