Date of Award

Spring 1-1-2012

Document Type


Degree Name

Master of Arts (MA)



First Advisor

Martha Palmer

Second Advisor

Laura Michaelis-Cummings

Third Advisor

James Martin


This paper describes a method of automatically comparing syntactic frames from the verb lexicon VerbNet with syntactic frames from the Semlink corpus. A method of extracting syntactic frames and semantic argument structures is explained, followed by a method of comparing syntactic frames, both directly and by argument structure. The results of the comparison are described in terms of matching success for frame tokens and frame types, divided into categories based on frame type frequency within Semlink. Overall, 54.14% of the frame tokens within Semlink can be directly matched to VerbNet, with an additional 14.32% matching by argument structure. However, only 29.30% of the frame types within Semlink can be matched to VerbNet, suggesting that the comparison method cannot match a majority of the large variation of frames types in Semlink. A set of distinguishing frame types for VerbNet classes is also proposed and included in this work.