Overview - Customizing models

Before you begin

1. If you haven't done so already, get started with Natural Language Understanding.
2. Check Language support for custom models to make sure that the custom model you want to create is supported.
3. Follow the customization instructions for one of the following features.
   • Entities and relations
   • Categories (Beta)
   • Classifications

Language support for custom models

Check the Custom model support columns in the tables on the Language support page to see the features that support custom models for each language.

Specifying training parameters for custom models

As part of the request to create or update a custom model, you may optionally include a training parameters object that specifies attributes of the model. For details, see feature-specific Classifications training parameters.

Targeted sentiment for custom model entities

For English only, you can get sentiment scores for each custom model entity that is detected by the service by setting the sentiment: true option in the entities object. No other languages support targeted sentiment for custom model entities.

Usage restrictions for custom models

The maximum number of models that can be trained via the Natural Language Understanding customization API in parallel is 3.