Mova0.5 pre-release⌕
☰
home / docs / treebanks / hr / UD_Croatian-SET

UD_Croatian-SET

Treebank card from UD release 2.18: metadata and texts come from UD_Croatian-SET/README.md; size is counted by streaming the .conllu files and checked against the release stats.xml. Generated by tools/treebanks build; do not edit by hand. Papers: Papers: UD_Croatian-SET; seeds: seeds/; converter: convert.md.

FieldValue
UD nameUD_Croatian-SET
LanguageCroatian (hr)
FamilyIE
Treebank codehr_set
UD pageuniversaldependencies.org/treebanks/hr set/index.html
LicenseCC BY-SA 4.0
Genresnews, web, wiki
Text in datayes
Parallelset
In UD sinceUD v1.1
AuthorsTanja Samardžić; Aleksandra Miletić; Nikola Ljubešić; Željko Agić; Daniel Zeman
Contacttanja.samardzic@uzh.ch
Developmentoutside the UD repo (elsewhere): fixes go to the source or the converter

Annotation layers

LayerOrigin
Lemmasconverted from manual
UPOSconverted from manual
XPOSmanual native
Featuresconverted from manual
Relationsmanual native

UD scale: manual native — annotated by hand directly in UD; converted from manual — converted from manual annotation in another scheme; converted with corrections — converted and partly corrected by hand; automatic with corrections — automatic, with partial manual corrections; automatic — automatic; not available — the layer is absent.

Size

SplitSentencesTokensWordsMultiword tokensEmpty nodes
train6 914152 857152 85700
dev96022 29222 29200
test1 13624 26024 26000
total9 010199 409199 40900

Tokens are surface tokens (Spanish vámonos is one), words are syntactic words (vamos + nos); multiword tokens are those split into several words; empty nodes are the N.M nodes of the enhanced graph.

Summary

The Croatian UD treebank is based on the extension of the SETimes-HR corpus, the hr500k corpus.

Introduction

The sentences are partially parallel with the smaller Serbian UD treebank, which comes from the Serbian edition of SETimes. For the UD release 2.4, the Croatian and Serbian corpus were enriched with newdoc metadata and re-split so that the corresponding documents are in the same section (train/dev/test) in Croatian and Serbian.

Sentence ids have also been changed to reflect the domain / source the data comes from, and not the dependence to one of the sections (train/dev/test), as was the case in previous releases.

Training set.

Contains 6,844 sentences (151,226 tokens) from three sources:

  1. sentence ids set.hr*: pseudorandom 80% of documents of newspaper text from the Southeast European Times news website, obtained from the SETimes parallel corpus. This part of the treebank is built on top of the SETimes.HR dependency treebank of Croatian;
  2. sentence ids news.hr*: pseudorandom 80% of documents of Croatian news web sources.
  3. sentence ids web.hr* : pseudorandom 80% of sentences of Croatian web sources.

Development set.

Contains 954 sentences (21,952 tokens) from three sources:

  1. sentence ids set.hr*: pseudorandom 10% of documents of newspaper text from the Southeast European Times.
  2. sentence ids news.hr*: pseudorandom 10% of documents of Croatian news web sources.
  3. sentence ids web.hr* : pseudorandom 10% of sentences of Croatian web sources.

Test set.

Contains 1214 sentences (26,263 tokens) from four sources:

  1. sentence ids set.hr*: pseudorandom 10% of documents of newspaper text from the Southeast European Times (+ the previous test set from the same source).
  2. sentence ids wiki.hr*: old Wikipedia-based test set.
  3. sentence ids news.hr*: pseudorandom 10% of documents of Croatian news web sources.
  4. sentence ids web.hr* : pseudorandom 10% of sentences of Croatian web sources.

Details

Sentence and word segmentation was manually checked. The treebank does not include multiword tokens. No language-specific features and relations were used. The POS tags and features were converted from Multext East v6 (present in the XPOS column) and manually checked. The syntactic annotation was done manually.

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/treebanks/hr/set/