$G=$;NO@J8!&2]Bj8&5f(B $BEE;R%U%!%$%k$HF1$8>l=j$KCV$$$F2<$5$$!#$h$m$7$/$*4j$$$7$^$9!#(B ->
The study of predicate-argument structure analysis has focused on verbs and adjectives. However, to fully understand the sentence ``His decision was right,'' both the predicate ``was right'' and verbal noun ``decision'' need to be identified as event denoting expressions. Here, we call verbal nouns event nouns, which refer to the event like ``he decided something,'' and propose argument structure analysis of event nouns. Some event nouns, such as ``cook'', do not always refer to an event and word sense disambiguation of event nouns has to be done depending on event nouns in use.
In this thesis, we present a machine learning method to identify eventness using noun-verb co-occurrences and noun phrase patterns from large corpora. We also investigate a method of incorporating noun-verb co-occurrences to label arguments of event nouns, and study errors for the argument labeling task.