TajikNLPWorld/tajik-sentiment-lexicon

🇹🇯 Tajik Sentiment Lexicon

A comprehensive sentiment lexicon for the Tajik language, containing 3,517 words annotated with polarity (positive/negative) and intensity scores.

📖 Description

This dataset provides a curated list of Tajik words with sentiment annotations:

• polarity: either позитив (positive) or негатив (negative)
• intensity: a floating-point score from -1.0 (most negative) to +1.0 (most positive), indicating the strength of sentiment.

The lexicon was created by merging two high-quality word lists: a positive dictionary and a negative dictionary, each with intensity ratings. Russian translations are included for each word, making this resource useful for cross-lingual research.

Applications

• Sentiment analysis for Tajik text
• Opinion mining and emotion detection
• Cross-lingual transfer learning (Tajik-Russian)
• Lexicographic studies

📊 Dataset Statistics

Overall Metrics

Metric	Value
Total Words	3,517
Positive Words	2,012
Negative Words	1,505
Avg. Positive Intensity	0.766
Avg. Negative Intensity	-0.727

Intensity Distribution (absolute values)

Intensity Range (abs)	Count
0.0-0.2	39
0.2-0.4	53
0.4-0.6	802
0.6-0.8	1658
0.8-1.0	965

📁 Data Format

Each record is a JSON object with the following fields:

Field	Type	Description	Example
tajik	string	The word in Tajik (Cyrillic script)	"хуб"
translation	string	Russian translation(s)	"хороший, добрый"
polarity	string	Sentiment polarity: позитив or негатив	позитив
intensity	float	Intensity score from -1.0 to +1.0	0.9

Example Record

{
  "tajik": "хуб",
  "translation": "хороший, добрый",
  "polarity": "позитив",
  "intensity": 0.9
}

🚀 Usage

Load with 🤗 Datasets

from datasets import load_dataset

dataset = load_dataset("TajikNLPWorld/tajik-sentiment-lexicon")
train_data = dataset["train"]

# Filter positive words
positive = train_data.filter(lambda x: x["polarity"] == "позитив")
print(f"Positive words: {len(positive)}")

Use with pandas

import pandas as pd
df = dataset["train"].to_pandas()
df.head()

📜 License

Apache License 2.0

🤝 Citation

@dataset{tajik_sentiment_lexicon,
    title = {Tajik Sentiment Lexicon},
    author = {TajikNLPWorld},
    year = {2025},
    publisher = {Hugging Face},
    url = {https://huggingface.co/datasets/TajikNLPWorld/tajik-sentiment-lexicon}
}

---

Generated with ❤️ for the Tajik NLP community