Talk by Dr. Joshi on Penn Tree Bank
Penn Tree Bank -
corpora of annotated text.
Penn Discourse Treebank (PDTB)
PDTB provides annotations of both lexicaly trigged discourse relations as well as inferred discourse relations triggered via structurally adjacency.
What is annotated
- Discourse Relations and their arguments
PTDB Annotation Overview
1) Explicit Connectives
2) Alternative LExicalizations
3) Implicit Conn
4) Entity Based Coherence Relation
5) No Relation (NoRel)
just because a connective appears it might not be necessary a discourse connective.
No Contraints on relative order of Arg1 and Arg2.
2) Non Linear