delph-in / delphin.redwoods

PyDelphin interface for the LinGO Redwoods Treebank

delphin.redwoods

Pydelphin plugin for the LinGO Redwoods Treebank

Installation

pip install delphin.redwoods

requirements:

• pydelphin >= 1.0.0
• svn >= 0.3.46

Usage

Treebank class that acts as a bundle, grouping testsuites of interest

from delphin.redwoods import Treebank

Redwoods data can be retrieved from 3 sources:

• User specified path to gold parse

profile = Treebank("wsj00a", "path/to/gold/")

• If environment $LOGONROOT is setup, use its remote copy of Redwoods

profile = Treebank("wsj00a")

• retreave svn repository to ~/redwoods<TAG> (default option)

profile = Treebank("wsj00a")

When retreiving the data you can specified which tag version to use. Currently supported versions:

• 1214 (default)
• 2018

User can edit the profiles stored in the bundle:

profile = Treebank("wsj00") # access to profile wsj00
profile.upload("wsj01") # access to profiles wsj00 and wsj01
profile.remove("wsj00") # access to profiles wsj01

There exists support to create standard bundle sets of profiles:

deepbank_train = Treebank("deepbank.train") # wsj section 0 - 19
deepbank_dev = Treebank("deepbank.dev") # wsj section 20
deepbank_test = ReTreebankdwoods("deepbank.test") # wsj section 21

TreebankResponse can be retrieved for a single profile or all profiles specified in the bundle:

profile.get("wsj00") # ``TreebankResponse`` for Profile wsj00
profile.get_all() # ``TreebankResponse`` for all profiles specified in ``profile``

Each TreebankResponse consists of the following information:

• metadata about the the partition, including its description, split, and stats.
• results for each sentence the following information is recorded:
  • surface: surface form
  • derivation: derivation information
  • tree: syntactic tree
  • mrs : mrs representation