Abstract
This chapter provides a summary of the available annotated resources for training supervised approaches to automatic semantic role labelling, and their theoretical underpinnings. It compares three main types of semantic role labelled resources: the PropBank, VerbNet, and FrameNet. It also surveys the different techniques that have been used to build supervised systems, as well as less supervised approaches. It examines the syntactic representations, features, and classifiers that are typically used in semantic role labelling systems. It concludes with a discussion of other languages, and issues that need to be considered when applying semantic role labelling cross-linguistically. It also discusses issues with the alignment of semantic roles in two languages and the projection of semantic roles from one language to another.
