---
language:
- tg
license: apache-2.0
tags:
- part-of-speech
- pos-tagging
- tajik-language
- lexical-resource
- nlp
datasets:
- tajik-pos-corpus
---

# 🇹🇯 Tajik POS Corpus

A comprehensive part-of-speech tagged lexicon for the Tajik language, containing over **52,508** unique word forms with their morphological categories.

## 📖 Description

This dataset is a curated collection of Tajik words annotated with part-of-speech tags according to standard Tajik grammatical classification. It includes both general vocabulary and proper nouns (including toponyms), and serves as a fundamental resource for:
- Training and evaluating POS taggers for Tajik
- Morphological analysis and lemmatization
- Lexicographic research
- Language learning applications

The corpus has been carefully cleaned: duplicates removed, artifacts (like `_1` suffixes) eliminated, and enclitics category verified against academic grammar.

## 📊 Dataset Statistics

### Overall Metrics
| Metric | Value |
|--------|-------|
| Total Records | 52,508 |
| Unique Tajik Words | 52,447 |
| Uniqueness Ratio | 99.88% |
| Unique POS Tags | 28 |

### Word Length Statistics
| Metric | Value |
|--------|-------|
| Average Length | 7.85 characters |
| Median Length | 7 |
| Minimum Length | 1 |
| Maximum Length | 70 |

### Part-of-Speech Distribution
| Part of Speech   | Count   | Percentage   |
|:-----------------|:--------|:-------------|
| исм              | 23,730  | 45.19%       |
| сифат            | 15,437  | 29.40%       |
| исми хос         | 8,894   | 16.94%       |
| зарф             | 1,525   | 2.90%        |
| феъл             | 1,345   | 2.56%        |
| нидо             | 248     | 0.47%        |
| шумора           | 139     | 0.26%        |
| символ           | 102     | 0.19%        |
| ихтисора         | 100     | 0.19%        |
| рақам            | 100     | 0.19%        |
| пайвандак        | 86      | 0.16%        |
| пешоянд          | 72      | 0.14%        |
| пасванд          | 70      | 0.13%        |
| ҷонишин          | 68      | 0.13%        |
| феъли ёридиҳанда | 63      | 0.12%        |
| замони феъл      | 62      | 0.12%        |
| ибора            | 61      | 0.12%        |
| ҳиссача          | 53      | 0.10%        |
| пешванд          | 51      | 0.10%        |
| воҳиди ченак     | 50      | 0.10%        |
| муайянкунанда    | 44      | 0.08%        |
| нишондиҳанда     | 40      | 0.08%        |
| модалӣ           | 39      | 0.07%        |
| тақлидӣ          | 37      | 0.07%        |
| пайванди табей   | 30      | 0.06%        |
| пайванди бабей   | 26      | 0.05%        |
| пасоянд          | 23      | 0.04%        |
| энклитика        | 13      | 0.02%        |

### 🔝 Top 10 Longest Words
|   # | Word                                                                   | POS      |   Length |
|----:|:-----------------------------------------------------------------------|:---------|---------:|
|   1 | Боғи фарҳангӣ-фароғатии марказии Кӯлоб ба номи Мир Сайид Алии Ҳамадонӣ | исми хос |       70 |
|   2 | Осорхонаи мероси хаттӣ ва адабии ба номи Мир Сайид Алии Ҳамадонӣ       | исми хос |       64 |
|   3 | Театри давлатии академии опера ва балети ба номи Садриддин Айнӣ        | исми хос |       63 |
|   4 | Филиали Донишгоҳи байналмилалии сайёҳӣ ва соҳибкории Тоҷикистон        | исми хос |       63 |
|   5 | Осорхонаи санъати миллии Тоҷикистон ба номи Камолиддин Беҳзод          | исми хос |       61 |
|   6 | Кумитаи иҷроияи Ҳизби Халқии Демократии Тоҷикистон дар Кӯлоб           | исми хос |       60 |
|   7 | Масҷиди марказии ҷомеи Душанбе ба номи Абуҳанифа Имоми Аъзам           | исми хос |       60 |
|   8 | Театри давлатии академии драмавии ба номи Абулқосим Лоҳутӣ             | исми хос |       58 |
|   9 | Коллеҷи санъати ҷумҳуриявии ба номи Кароматулло Қурбонов               | исми хос |       56 |
|  10 | Театри давлатии академии опера ва балети ба номи С. Айнӣ               | исми хос |       56 |

### 📝 First Letter Distribution (Top 10)
| First Letter   | Count   | Percentage   |
|:---------------|:--------|:-------------|
| б              | 4,303   | 8.19%        |
| м              | 4,152   | 7.91%        |
| с              | 3,353   | 6.39%        |
| н              | 2,486   | 4.73%        |
| д              | 2,376   | 4.53%        |
| т              | 2,301   | 4.38%        |
| а              | 2,273   | 4.33%        |
| к              | 2,105   | 4.01%        |
| п              | 2,076   | 3.95%        |
| х              | 1,743   | 3.32%        |

### 🔤 Special Words
- **Words with spaces** (multi-word expressions): 5 examples (total 5)
- **Words with hyphens**: 5 examples

## 📁 Data Format

Each record is a JSON object with two fields:

| Field | Type | Description | Example |
|-------|------|-------------|---------|
| `tajik` | string | The word form in Tajik (Cyrillic script) | `"Рунҷ"` |
| `part_of_speech` | string | The part-of-speech tag | `"исми хос"` |

### Example Record
```json
{
  "tajik": "Рунҷ",
  "part_of_speech": "исми хос"
}
```

## 🚀 Usage

### Load with 🤗 Datasets
```python
from datasets import load_dataset

dataset = load_dataset("TajikNLPWorld/tajik-pos-corpus")
train_data = dataset["train"]

# Filter nouns
nouns = train_data.filter(lambda x: x["part_of_speech"] == "исм")
print(f"Number of nouns: {len(nouns)}")
```

## 📜 License
Apache License 2.0

## 🤝 Citation
```
@dataset{tajik_pos_corpus,
    title = {Tajik POS Corpus},
    author = {Arabov Mullosharaf Kurbonovich, TajikNLPWorld},
    year = {2025},
    publisher = {Hugging Face},
    url = {https://huggingface.co/datasets/TajikNLPWorld/tajik-pos-corpus}
}
```

---
*Generated with ❤️ for the Tajik NLP community*