use Tree Adjoining Grammar derivations

[anoopsarkar]

add TAG elementary trees (etrees) in the edge features and also implement the method for building the derived tree described in the Carreras et al paper.

[yulanh]

Waiting for the data.

[anoopsarkar]

Let us start by using this data:

• LTAG spinal data
• LTAL spinal website

Note that this treebank does not obey the property used in the glm-parser paper. I will think about how to deal with that issue, but I think we can get started using this data. The first step will be to map this LTAG spinal treebank format into the conll format.

[anoopsarkar]

we now have a way to generate the LTAG spinal training data directly from the dependency trees and the Penn Treebank phrase structure files. however, we are waiting for cube pruning to be implemented to add the LTAG features into the parser.