Frame annotation

Example project (SRL/events)

This project provides a short introduction to modeling frame-like annotation tasks such as semantic role labeling or event annotation.

The two layers SemPred and SemArg are used to model the frame. First, a SemPred annotation is created on the words that trigger the frame. The arguments of the frame can then be linked up to this SemPred annotation by creating SemArg annotations on the words representing the arguments and linking them into the arguments feature of the SemPred annotation.

To facilitate the annotation, the project includes a small set of constraint rules. There are two types of rules. The first type controls the category feature of the SemPred annotation based on the lemma of the trigger word. The second type of rules controls the slots of the arguments feature of the SemPred layer. To allow the constraints to work, the category and arguments feature have each been associated with a tagset.