Refining Syntactic Categories Using Local Contexts — Experiments in Unlexicalized PCFG Parsing
Pate, John K.
MetadataShow full item record
As an extension of decades of syntactic theorizing, treebanks have inherited a small set of phrasal categories, which abstract over the environments that the categories can occur in. Extending ideas from Johnson (1998), we explore encoding information from the local tree context in each category. We then discuss two clustering techniques which preserve the distributionally relevant category distinction, forming linguistically relevant generalizations and improving PCFG parsing performance.