Use this model type to extract structure information from unstructured text. This model type takes annotated documents (from a labeling tool like teach.indico.io) and trains a custom collection to replicate how you’ve highlighted words and phrases in your training data. Train an annotation custom collection for custom named entity recognition, highlighting price mentions in text, or extracting positive and negative sentiment phrases from reviews.
Adding data for this type of model requires a set of character offsets bounding values from the start of the sample as shown below.
// Annotation Collections are currently not supported by the Java client library.