Paper: Fast Computing Grammar-driven Convolution Tree Kernel for Semantic Role Labeling

ACL ID I08-2109
Title Fast Computing Grammar-driven Convolution Tree Kernel for Semantic Role Labeling
Venue International Joint Conference on Natural Language Processing
Session Main Conference
Year 2008
Authors

Grammar-driven convolution tree kernel (GTK) has shown promising results for se- mantic role labeling (SRL). However, the time complexity of computing the GTK is exponential in theory. In order to speed up the computing process, we design two fast grammar-driven convolution tree kernel (FGTK) algorithms, which can compute the GTK in polynomial time. Experimental re- sults on the CoNLL-2005 SRL data show that our two FGTK algorithms are much faster than the GTK.