资 源 简 介
Introduction
This package contains syntax-tree conversion software developed in
conjunction with a multi-layered and multi-representational treebank
for Hindi and Urdu. The purpose of this software is (a) to learn
mapping rules from dependency-structure (DS) to phrase-structure (PS)
given a training corpus and (b) to apply them to unseen DS trees,
generating the corresponding PS representation. Ultimately the goal
of this software is to generate the phrase structure layer for the
entire treebank, given the DS and other hand-annotated layers
(e.g. PropBank) which prove useful in improving the accuracy of the
yielded trees.
Necessary inputs:
* training data: a set of compatible dependency and phrase structure tree pairs
* test data: dependency structure trees for conversion
Optional inputs:
* propbank (or other markup) that has been annotated on the same training data
Outputs:
* from training data: rules for converting dependency to phrase structu
