Computational Grammars: Treebanking

We will use the Full Forest Treebanker (fftb)

• Select from a packed forest (all parses)
• 17% more decisions (on average, about one per item, depending on complexity)
• But fewer costly rejections
• New web interface

Definitive documentation on line at the DELPH-IN wiki: FFTB

• Parse a profile (with the packed forest)
• If you have a gold profile then update
• Otherwise just treebank
• You can also train a model

Treebank your testsuite and corpus then commit them to github
Normally in tsdb/gold/profile

Back to course page

Course materials borrow heavily from Linguistics 567: Knowledge Engineering for NLP at the University of Washington. Thanks to Emily Bender for letting us use them.