UzEDSA

Dataset Summary

UzEDSA is a large-scale Uzbek dataset for emotion detection and sentiment analysis. The current release contains 304,364 annotated rows in a single TSV file, covering texts from nine domains.

Each row includes:

• a free-form Uzbek text snippet
• a primary emotion label
• a multi-label emotion annotation (semicolon-separated)
• a 3-way sentiment polarity label
• an emotion intensity score
• a sarcasm flag
• a domain tag

The annotation scheme follows Plutchik's eight basic emotions extended with a neutral and a mixed category.

Supported Tasks

• emotion classification (single-label)
• multi-label emotion detection
• sentiment classification
• sarcasm detection

Files

• UzEDSA_dataset.tsv: main file in tab-separated format

Dataset Structure

Columns

Column	Type	Description
№	integer	Unique sequential identifier
Text_uz (izoh)	string	Uzbek text or comment
Emotion_primary	categorical	Dominant emotion label (one of 10 classes)
Emotion_multi	string	All applicable emotion labels, semicolon-separated
Polarity	categorical	Sentiment polarity: POS, NEG, NEU, or MIX
Intensity	integer	Emotion intensity on a scale of 0 to 2
Sarcasm	binary	0 or 1, whether sarcasm or irony is present
Domain	categorical	Thematic domain of the text

Tagsets

Emotion Labels

Label	Full name	Description	Example
JOY	Joy	Happiness, satisfaction, gratitude	Shifokorga katta rahmat!
NEU	Neutral	Factual or informational statement	Qabul ertaga 9:00 da boshlanadi.
ANG	Anger	Rage, irritation, frustration	Bu qanday xizmat o'zi!
SAD	Sadness	Grief, disappointment, low mood	Ahvolim yana yomonlashdi.
ANT	Anticipation	Hope, expectation, looking forward	Yangi kursdan yaxshi natija kutyapman.
DIS	Disgust	Repulsion, aversion, loathing	Xonalar iflos, ko'rish ham jirkanch!
SUR	Surprise	Astonishment, unexpected outcome	Bunaqa tez tiklanaman deb o'ylamagan edim.
MIX	Mixed	Two or more contradictory emotions	Navbat uzoq, hafaman, lekin shifokorlar yaxshi.
TRU	Trust	Confidence, reliability, faith	Doim shu shifokorga ishonaman.
FEA	Fear	Anxiety, worry, dread	Analiz natijasini kutib xavotirdan uxlay olmadim.

Polarity Labels

Label	Meaning	Description	Example
POS	Positive	Favourable or beneficial experience	Operatsiya muvaffaqiyatli o'tdi!
NEG	Negative	Unfavourable, harmful, or disappointing	Bu doridan umuman foyda bo'lmadi.
NEU	Neutral	Factual or emotionally flat statement	Bugun yangi apparat ishga tushdi.
MIX	Mixed	Contradictory positive and negative signals	Navbatdan charchadim, lekin natija yaxshi bo'ldi.

Intensity Scale

Score	Meaning	Description	Example
0	None / Neutral	No emotional charge	Ertaga tahlil bor.
1	Moderate	Mild emotional expression	Og'riqlar kamaydi, yaxshi.
2	Strong	Intense emotional expression	Ajoyib natija, juda xursandman!

Domain Labels

Domain	Description	Example context
general	General everyday topics	Everyday life comments and opinions
personal	Personal experiences and feelings	Private thoughts, personal situations
news	News reports and public events	Media-based reactions, current affairs
tech	Technology and digital services	Apps, devices, online platforms
med	Medical and healthcare	Hospitals, doctors, symptoms, medicines
edu	Education	Schools, courses, learning experiences
transport	Transport and commuting	Buses, taxis, traffic, schedules
service	Service and administrative encounters	Clinics, call centres, offices, queues
marketing	Advertising and commercial content	Product reviews, promotional messages

Statistics

Overview

• Rows: 304,364
• Columns: 8
• Split: train only

Emotion Distribution (Emotion_primary)

Label	Count
JOY	104,783
NEU	98,382
ANG	25,731
SAD	23,999
ANT	20,922
DIS	15,391
SUR	6,041
MIX	4,008
TRU	2,631
FEA	2,476

Polarity Distribution

Label	Count
POS	134,377
NEU	98,382
NEG	67,597
MIX	4,008

Intensity Distribution

Score	Count
0	98,382
1	128,953
2	77,029

Sarcasm

Value	Count
0 (not sarcastic)	302,503
1 (sarcastic)	1,861

Domain Distribution

Domain	Count
general	198,438
personal	34,071
news	28,858
tech	10,956
med	7,455
edu	7,163
transport	6,505
service	5,911
marketing	5,007

Top Multi-label Emotion Combinations (Emotion_multi)

Combination	Count
JOY;ANT	6,835
ANG;DIS	1,872
SAD;ANT	1,731
ANT;JOY	1,651
JOY;SUR	1,629
ANG;JOY	1,096
DIS;ANG	1,047
SAD;JOY	852
JOY;TRU	794
ANG;SAD	600

Usage

from datasets import load_dataset

dataset = load_dataset(
    "csv",
    data_files={"train": "UzEDSA_dataset.tsv"},
    delimiter="\t",
    encoding="utf-8-sig",
)

print(dataset["train"][0])

Or directly from the Hub:

from datasets import load_dataset

dataset = load_dataset("uznlp-uz/UzEDSA")
print(dataset["train"][0])

Citation

If you use UzEDSA in your research, please cite this dataset.

License

This dataset is released under the CC-BY-4.0 license.