Video Recordings

Shallow semantic analyzers, such as semantic role labelers and sense taggers, are increasing in accuracy and becoming commonplace. However, they only provide limited and local representations of words and individual predicate-argument structures. This talk will address some of the current opportunities and challenges in producing deeper, richer representations of coherent eventualities. Available resources, such as VerbNet, that can assist in this process will also be discussed, as well as some of their limitations.