A mapping configuration from lexical units in Latvian FrameNet to PropBank frames, and a mapping configuration from FrameNet frame elements and their UD roles in Latvian UD Treebank roles to PropBank semantic roles for the given pair of FrameNet and PropBank frames.
An automatically derived Latvian PropBank annotation layer of the FullStack corpus. PropBank annotations are derived based on the mapping configurations.
Fields: [ID, FORM, LEMMA, UPOS, XPOS, FEATS, HEAD, DEPREL] + [FILLPRED, PRED, APRED]
- FILLPRED: indicates the target word
- PRED: semantic frame
- APRED: semantic roles
Note that each corpus example contains one PropBank annotation set. If there are several annotation sets for the same sentence, they are included as separate corpus examples.
- [Mapping] Target verbs: 576 out of 865 (67%)
- [Mapping] FrameNet frames: 255 out of 458 (56%)
- [Mapping] FrameNet lexical units (verb-frame pairs): 788 out of 1,660 (48%)
- [Mapping] PropBank frames (predicates): 459 out of 7,311 (6%)
- [Data] Annotation sets: 5,473 out of 10,951 (50%)