Overview to NLU modeling
Four main steps of NLU model building.
NLU modeling best practices
This section describes best practices for creating high-quality NLU models that can interpret the meaning of user text inputs.
This section is not meant to provide details about the mechanics of how to create an NLU model in Mix.nlu. Instead, it aims to provide a set of best practices for developing more accurate NLU models more quickly, from designing an ontology and creating a training set to evaluating and improving the model. The intended audience is developers with at least a basic familiarity with the Mix.nlu model development process.
For best practices on building models to support speech recognition, see DLM and ASR tuning best practices.
Designing the model
Best practices to follow when designing NLU models.
Generating data and training the initial model
Two approaches to gathering data for training, deployment usage data and artificial data.
Evaluating NLU accuracy
Best practices around generating test sets and evaluating NLU model accuracy.
Improving NLU accuracy
Important considerations for improving NLU accuracy.
FAQs
Frequently asked question around NLU modeling.