UzWordNet-GoldCore

UzWordNet-GoldCore is a gold-core Uzbek lexical-semantic dataset organized in a WordNet-style structure. It contains Uzbek lexical entries with synonyms, antonyms, definitions, part-of-speech labels, domain and ontology labels, example sentences, and semantic relations.

Dataset Summary

• Dataset ID: uznlp-uz/UzWordNet-GoldCore
• Language: Uzbek (uz)
• Rows: 11,178
• Columns: 14
• Split: train
• Format: UTF-8 TSV
• Data file: uzwordnet_synsets.tsv
• License: CC BY 4.0

Data Fields

Field	Description
id	Record identifier.
word	Uzbek lexical entry or lemma.
synonyms	Comma-separated synonym list.
antonyms	Antonym list when available.
definition	Uzbek definition or gloss.
pos	Part-of-speech label.
style	Register or style label.
domain	Domain label.
ontology_category	Ontology category label.
example	Example sentence when available.
hypernym	Hypernym label.
hyponym	Hyponym list when available.
holonym	Holonym list when available.
meronym	Meronym list when available.

Label Values

POS values: bogʻlovchi, feʼl, koʻmakchi, modal, olmosh, ot, ravish, sifat, son, undov, yuklama

Domain values: anatomiya/tibbiyot, botanika, din, huquq, kimyo, matematika, musiqa, texnika, tilshunoslik, umumiy, zoologiya

Statistics

Overview

• Rows: 11,178
• Columns: 14
• Split: train only
• Unique words: 6,506
• Unique definitions: 2,706
• Duplicate rows: 0
• Examples available: 11,169
• Examples missing: 9

POS Distribution

POS	Count
ot	4,774
sifat	3,476
feʼl	2,133
ravish	481
koʻmakchi	104
modal	65
olmosh	46
yuklama	44
undov	27
bogʻlovchi	23
son	5

Style Distribution

Style	Count
neytral	8,918
badiiy	1,256
soʻzlashuv	777
rasmiy	158
kitobiy	69

Domain Distribution

Domain	Count
umumiy	7,610
tilshunoslik	1,761
anatomiya/tibbiyot	504
matematika	367
zoologiya	348
botanika	240
huquq	104
texnika	77
musiqa	73
din	67
kimyo	27

Ontology Category Distribution

Ontology category	Count
belgi/xususiyat	2,591
mavhum tushuncha	2,182
harakat/jarayon	2,133
shaxs	714
tana qismi	601
rang belgisi	578
predmet/narsa	569
holat/daraja belgisi	481
joy/makon	284
hissiyot	282
oʻlcham belgisi	237
yordamchi soʻz	171
oʻsimlik	113
modal munosabat	65
baholovchi belgi	60
hayvon	55
hissiy ifoda	27
umumiy tushuncha	25
yorugʻlik belgisi	10

Hypernym Distribution

Hypernym	Count
belgi	3,476
mavhum tushuncha	2,182
harakat	2,133
shaxs	714
tana qismi	601
predmet/narsa	569
holat belgisi	481
joy/makon	284
hissiyot	282
yordamchi soʻz	171
oʻsimlik	113
modal munosabat	65
hayvon	55
hissiy ifoda	27
umumiy tushuncha	25

Relation Coverage

Field	Non-empty	Empty
synonyms	11,178	0
antonyms	1,935	9,243
hypernym	11,178	0
hyponym	58	11,120
holonym	40	11,138
meronym	19	11,159

Normalization

• Text is stored as UTF-8.
• Leading and trailing whitespace was removed.
• Repeated whitespace was collapsed to one space.
• Embedded tabs and newlines were removed so each record stays on one TSV row.
• Uzbek apostrophes were normalized to distinct Unicode characters: ʻ for oʻ/gʻ and ʼ for the tutuq sign.
• Cyrillic lookalike characters in Latin text were mapped to Latin equivalents.
• Optional relation fields are empty when no relation is available.
• Categorical values in pos, style, domain, ontology_category, and hypernym are lowercased.

Loading

from datasets import load_dataset

dataset = load_dataset("uznlp-uz/UzWordNet-GoldCore", split="train")
print(dataset[0])

The TSV file can also be loaded directly:

from datasets import load_dataset

dataset = load_dataset(
    "csv",
    data_files="uzwordnet_synsets.tsv",
    delimiter="\t",
    split="train",
)

Intended Use

This dataset can be used for Uzbek NLP research and development, including lexical-semantic analysis, synonym and antonym modeling, semantic relation extraction, dictionary-based applications, and evaluation of Uzbek language models.

Notes

List-like fields are stored as strings. Commas separate lexical items inside a group; semicolons separate grouped relation sets where present.

Some examples and optional semantic relation fields are empty because those values are not available for every lexical entry.

Citation

If you use UzWordNet-GoldCore, cite the dataset repository:

@misc{uzwordnet_goldcore,
  title = {UzWordNet-GoldCore},
  author = {{Elov B.B., Alaev R.H.}},
  year = {2026},
  howpublished = {\url{https://huggingface.co/datasets/uznlp-uz/UzWordNet-GoldCore}},
  license = {CC BY 4.0}
}