Note
Access to this page requires authorization. You can try signing in or changing directories.
Access to this page requires authorization. You can try changing directories.
Natural Language Understanding (NLU) is a branch of artificial intelligence that enables computers to comprehend and interpret human language. This understanding goes beyond simply processing words. Copilot Studio provides different NLU options, based on the needs of your application.
Generative AI
Generative AI is the default option and is used to respond with the best combination of actions, topics, and knowledge sources. Makers need to provide high-quality descriptions for all aspects of your agent, which then lets generative AI orchestrate the conversation for you.
The generative AI option is best suited for applications where you want to perform a minimal amount of setup, and you're comfortable with allowing generative AI to orchestrate the conversation. There might also be other costs associated with this option.
The generative AI option is configured in the agent's settings (Generative AI > Orchestration > Yes). For more information, go to Orchestrate agent behavior with generative AI.
Classic orchestration
If you'd prefer a more deterministic option for your application, then choose the "classic" Copilot Studio orchestration options in your agent's settings (Generative AI > Orchestration > No).
There are three "classic" options: NLU, NLU+, and CLU. All of them provide full, repeatable control over your agent's conversations using a customized dialog.
NLU
If you want an easier programmable design or have simpler orchestration needs, the original NLU option is useful. With this option, you can quickly add 5 to 20 short phrases per topic and create RegEx or List custom entities. You also don't need to add entity annotations within your training data.
Note
Latency might increase if you add too much training data. For more information, go to AI features for Teams and Classic chatbots.
NLU+
If you need to achieve high accuracy, then use the NLU+ option. The NLU+ option is ideal for large enterprise-grade applications. These types of applications typically consist of a large number of topics and/or entities, and use a large number of training samples. Also, if you have a voice-enabled agent, your NLU+ training data is also used to optimize your speech recognition capabilities.
Important
The NLU+ option is available when you manage your voice or chat channels with a Dynamics 365 Contact Center license. For more information, go to System requirements for Dynamics 365 Contact Center.
For the highest accuracy, add entity annotations to your topic trigger phrases. Also add training samples to illustrate how customers might respond to questions regarding specific custom entities.
With the NLU+ option, the model is precompiled, which helps ensure consistent performance, regardless of the volume of training data. To deploy the model in production, it must first be trained. This step enables the system to optimize for your specific use case while maintaining predictable performance at runtime.
Azure Conversational Language Understanding (CLU)
For makers with an Azure subscription and existing Azure models, you can link your CLU model to your agent and let the model drive conversations. However, this option requires an Azure subscription, management of the model in Azure, and maintenance to keep the model and agent synchronized. For more information, go to Conversational language understanding integration overview.