Varro: An Algorithm and Toolkit for Regular Structure Discovery in Treebanks

摘要

The Varro toolkit is a system for identi-fying and counting a major class of reg-ularity in treebanks and annotated nat-ural language data in the form of tree-structures: frequently recurring unordered subtrees. This software has been designed for use in linguistics to be maximally applicable to actually existing treebanks and other stores of tree-structurable nat-ural language data. It minimizes mem-ory use so that moderately large treebanks are tractable on commonly available com-puter hardware. This article introduces condensed canonically ordered trees as a data structure for efficiently discovering frequently recurring unordered subtrees.