Original PropBank is a corpus (PennTreeBank), annotated with predicates (think of them as verbal propositions), their arguments and their relation to the predicate. That relation is referred to as a semantic role and this task is named as semantic role labeling (SRL), a.k.a shallow semantic parsing.
A sample sentence annotated with semantic roles accordingly in FrameNet, VerbNet and PropBank convention is shown below. Unlike FrameNet and VerbNet, PropBank (PB) does not make use of a reference ontology like semantic frames or verb classes. Instead, semantic roles are numbered from Arg0 to Arg5 for the core arguments.
Recent advances in Semantic Role Labeling and its potential usages on various NLP applications like information retrieval, text summarization, question answering systems and text simplification has motivated us to create a Turkish PropBank. More details/resources can be found in our project website.