CCG Supertagging as Top-down Tree Generation
- Jakob Prange (Georgetown University)
- Nathan Schneider (Georgetown University)
- Vivek Srikumar (University of Utah)
Abstract
Although current CCG supertaggers achieve high accuracy on the standard WSJ test set, few systems make use of the categories\' internal structure that will drive the syntactic derivation during parsing. The tagset is traditionally truncated, discarding the many rare and complex category types in the long tail. Rather than give up on rare tags, we investigate models that account for the internal structure of categories, including novel methods for tree-structured prediction. Our best tagger is capable of recovering a sizeable fraction of long-tail supertags and even generates CCG categories that have never been seen in training, while approximating the prior state of the art in overall tag accuracy with fewer parameters.
Keywords: CCG, supertagging, long tail, structured prediction, syntax, robustness, tree decoding
How to Cite:
Prange, J., Schneider, N. & Srikumar, V., (2021) “CCG Supertagging as Top-down Tree Generation”, Society for Computation in Linguistics 4(1), 351-354. doi: https://doi.org/10.7275/s7gd-5n83
Downloads:
Download PDF