Summary

Simplified Chinese Universal Dependencies dataset converted from the GSD (traditional) dataset with manual corrections.

Introduction

This is a simplified Chinese version of the UD Chinese GSD treebank. It is initially automatically converted into simplified Chinese with the OpenCC tool with patterns for mapping punctuation, then corrected with manual fixes.

Changelog

• 2023-11-15 v2.13
  • Some PART/ADV should be SCONJ (see https://github.com/UniversalDependencies/docs/issues/460 and https://github.com/slavpetrov/parallel-treebanks/issues/30).
  • Fixed attachment clf+nummod according to the guidelines.
• 2023-05-15 v2.12
  • Fixed: PUNCT nodes must be attached via punct relations.
  • Fixed: Only some UPOS categories are compatible with mark.
  • Fixed: Only some UPOS categories are compatible with det.
  • Fixed: ADJ cannot be copula.
  • Fixed: Auxiliary must be tagged AUX.
  • Fixed: Nominal cannot be advmod.
  • Fixed: Verb cannot be advmod.
  • Added pinyin transcription.
  • Scaled down the set of copulas and other auxiliaries.
  • Fixed: function words must be leaves.
  • Fixed: case marker adpositions mislabled as acl.
• 2021-05-15 v2.8
  • Changed mark:relcl to mark:rel (as in the other Chinese treebanks).
  • Removed the relation case:dec (for 的 between two nouns; the other treebanks use just case here).
  • Removed the relation aux:aspect (the aspect particles 了 (le), 过 (guo), 着 (zhe) use just aux in the other treebanks).
  • Question particles changed from Mood=Inter to PartType=Int, and from discourse to discourse:sp.
  • Undocumented relation subtypes case:pref and case:suff changed to case.
  • Extent constructions converted from cop + mark:comp to compound:ext.
  • Changed mark:advb to mark:adv (as in the other Chinese treebanks).
• 2020-11-15 v2.7
  • Aspect markers relations are corrected from case:aspect to aux:aspect.
• 2019-11-15 v2.5
  • Initial release in Universal Dependencies, converted from UD_Chinese-GSD.
  • Google gave permission to drop the "NC" restriction from the license. This applies to the UD annotations (not the underlying content, of which Google claims no ownership or copyright).
  • Fixed punctuation (use East Asian punctuation where appropriate)
  • Fixed various parses and features (e.g., added Case=Ord)
  • Some manual fixes in tokenization

=== Machine-readable metadata (DO NOT REMOVE!) ================================
Data available since: UD v2.5
License: CC BY-SA 4.0
Includes text: yes
Genre: wiki
Lemmas: automatic with corrections
UPOS: converted with corrections
XPOS: manual native
Features: automatic with corrections
Relations: converted from manual
Contributors: Qi, Peng; Yasuoka, Koichi
Contributing: here
Contact: pengqi@cs.stanford.edu
===============================================================================