README.md

---
language:
- tg
- fa
license: apache-2.0
tags:
- lexical-corpus
- tajik-persian
- parallel-corpus
- dictionary
- bilingual
- cross-lingual
- linguistic-resource
datasets:
- TajPersLexicalCorpus
task_categories:
- translation
- token-classification
- fill-mask
- text-generation
- sentence-similarity
- feature-extraction
---

# TajPersParallelLexicalCorpus

A parallel lexical corpus for Tajik-Persian language pairs with usage examples and part-of-speech tagging.

## 📖 Description
This dataset contains parallel lexical pairs between Tajik and Persian languages, including usage examples and part-of-speech annotations. It serves as a valuable resource for linguistic research, machine translation, and natural language processing tasks involving Tajik and Persian languages.

## 👤 Author
**Arabov Mullosharaf Kurbonovich**  
*Creator and maintainer of the TajikPersianParallelLexicalCorpus*

## 📊 Dataset Statistics

### Overview
| Metric | Value |
|--------|-------|
| Total Records | 43,819 |
| Unique Tajik Words | 43,218 |
| Unique Persian Words | 40,435 |
| Total Examples | 22,635 |
| Average Examples per Record | 0.52 |

### Data Quality
| Metric | Count |
|--------|-------|
| Records with Empty Tajik | 0 |
| Records with Empty Persian | 1,753 |
| Records with Empty POS | 64 |

### Part of Speech Distribution
| Part of Speech (Tajik) | English Translation | Count | Percentage |
|:----------------------|:-------------------|------:|-----------:|
| исм | Noun | 24,179 | 55.18% |
| сифат | Adjective | 15,609 | 35.62% |
| зарф | Adverb | 1,548 | 3.53% |
| феъл | Verb | 1,391 | 3.17% |
| исми хос | Proper Noun | 443 | 1.01% |
| нидо | Interjection | 246 | 0.56% |
| шумора | Numeral | 146 | 0.33% |
| пайвандак | Conjunction | 91 | 0.21% |
| ҷонишин | Pronoun | 38 | 0.09% |
| ҳиссача | Particle | 30 | 0.07% |
| пешоянд | Preposition | 25 | 0.06% |
| пасоянд | Postposition | 9 | 0.02% |
| **TOTAL** | | **43,819** | **100%** |

## 📁 Data Format
Each record contains the following fields:

| Field | Type | Description |
|-------|------|-------------|
| `tajik` | string | Word/phrase in Tajik language (Cyrillic script) |
| `persian` | string | Corresponding word/phrase in Persian language (Arabic script) |
| `part_of_speech` | string | Part of speech tag in Tajik (e.g., "исм" for noun, "феъл" for verb) |
| `examples` | list[string] | Usage examples with poetic and literary references from classical and modern Persian/Tajik literature |
| `_queried_word` | string | Original queried word (internal field used during data collection) |

## 🎯 Intended Uses

This dataset is designed for the following NLP tasks:

| Task Category | Description | Example Use Case |
|--------------|-------------|------------------|
| **translation** | Word-level translation between Tajik and Persian | Building a bilingual dictionary |
| **token-classification** | Part-of-speech tagging for Tajik and Persian | Training a POS tagger |
| **fill-mask** | Masked language modeling for bilingual contexts | Pretraining cross-lingual models |
| **text-generation** | Language modeling for Tajik and Persian | Generating Tajik/Persian text |
| **sentence-similarity** | Cross-lingual word embedding alignment | Learning word representations |
| **feature-extraction** | Extracting word features for downstream tasks | Creating word vectors |

## 🚀 Usage Examples

### Load the dataset
```python
from datasets import load_dataset

# Load the dataset
dataset = load_dataset("TajikNLPWorld/TajPersParallelLexicalCorpus")

# Access the training split
train_data = dataset["train"]

# View first record
print(train_data[0])

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

# Get statistics
print(f"Dataset size: {len(train_data)} records")
print(f"Unique Tajik words: {len(set(train_data['tajik']))}")
```

### Example Record
```json
{
    "tajik": "модар",
    "persian": "مادر",
    "part_of_speech": "исм",
    "examples": ["Ҷони киромӣ ба падар боздод, Қолбади тира ба модар супурд. - Рӯдакӣ"]
}
```

### Translation Task
```python
# Create a Tajik-Persian translation dictionary
translation_pairs = []
for item in train_data:
    if item['tajik'] and item['persian']:
        translation_pairs.append({
            'tajik': item['tajik'],
            'persian': item['persian']
        })

print(f"Created {len(translation_pairs)} translation pairs")

# Example usage
word = "модар"
matches = [pair for pair in translation_pairs if pair['tajik'] == word]
if matches:
    print(f"{word} -> {matches[0]['persian']}")
```

### Token Classification (POS Tagging)
```python
# Prepare data for POS tagging
pos_data = []
for item in train_data:
    if item['part_of_speech'] and item['tajik']:
        pos_data.append({
            'tokens': [item['tajik']],
            'pos_tags': [item['part_of_speech']]
        })

print(f"Prepared {len(pos_data)} examples for POS tagging")

# Analyze POS distribution
from collections import Counter
pos_counts = Counter([item['pos_tags'][0] for item in pos_data])
for pos, count in pos_counts.most_common(5):
    print(f"{pos}: {count}")
```

### Fill-Mask / Language Modeling
```python
# Prepare text for language modeling
lm_texts = []
for item in train_data:
    # Add word pairs
    if item['tajik'] and item['persian']:
        lm_texts.append(f"таджикӣ: {item['tajik']} форсӣ: {item['persian']}")
    
    # Add examples
    for example in item['examples']:
        if example:
            lm_texts.append(example)

print(f"Prepared {len(lm_texts)} texts for language modeling")
print(f"Sample: {lm_texts[0]}")
```

### Sentence Similarity / Feature Extraction
```python
# Create pairs for similarity learning
similar_pairs = []
for i, item in enumerate(train_data[:1000]):  # Limit for example
    if item['tajik'] and item['persian']:
        # Similar pair (same meaning)
        similar_pairs.append({
            'sentence1': item['tajik'],
            'sentence2': item['persian'],
            'label': 1.0  # similar
        })
        
        # Different pair (different meaning)
        if i > 0 and train_data[i-1]['tajik']:
            similar_pairs.append({
                'sentence1': item['tajik'],
                'sentence2': train_data[i-1]['tajik'],
                'label': 0.0  # not similar
            })

print(f"Created {len(similar_pairs)} pairs for similarity learning")
```

## 🔬 Research Applications

| Research Area | How to Use This Dataset |
|--------------|------------------------|
| **Computational Linguistics** | Study lexical similarities between Tajik and Persian |
| **Machine Translation** | Build and evaluate word-level translation models |
| **Cross-lingual NLP** | Develop multilingual word embeddings and representations |
| **Morphological Analysis** | Analyze word formation patterns in both languages |
| **Lexical Semantics** | Study semantic relations and word meanings |

## 🌍 Languages
- **Tajik (tg)**: Cyrillic script, variant of Persian spoken in Tajikistan
- **Persian (fa)**: Arabic script, also known as Farsi, spoken in Iran, Afghanistan, and Tajikistan

## 📊 Dataset Creation
This dataset was created by:
1. Collecting parallel lexical pairs from various Tajik-Persian linguistic resources
2. Annotating parts of speech for each entry
3. Adding usage examples from classical and modern literature
4. Deduplicating and cleaning the data
5. Validating the quality of translations and annotations

## 📜 License
This dataset is released under the Apache 2.0 License. You are free to:
- Use the dataset for commercial and non-commercial purposes
- Modify and adapt the dataset
- Distribute copies and derivatives
- Use it for research and development

## 🤝 Citation
If you use this dataset in your research, please cite:

**APA:**
```
Arabov, M. K. (2026). TajPersParallelLexicalCorpus: A Parallel Lexical Corpus for Tajik-Persian [Data set]. Hugging Face. https://huggingface.co/datasets/TajikNLPWorld/TajPersParallelLexicalCorpus
```

**BibTeX:**
```bibtex
@dataset{arabov_tajpers_2026,
    title = {TajPersParallelLexicalCorpus: A Parallel Lexical Corpus for Tajik-Persian},
    author = {Arabov, Mullosharaf Kurbonovich},
    year = {2026},
    publisher = {Hugging Face},
    url = {https://huggingface.co/datasets/TajikNLPWorld/TajPersParallelLexicalCorpus}
}
```

## 📬 Contact
**Author:** Arabov Mullosharaf Kurbonovich  
For questions, collaborations, or contributions:
- Hugging Face: [TajikNLPWorld](https://huggingface.co/TajikNLPWorld)
- GitHub: [TajikNLPWorld](https://github.com/TajikNLPWorld)
- Email: [cool.araby@gmail.com]

## 🙏 Acknowledgments
Special thanks to:
- Contributors to Tajik and Persian lexical resources
- The Hugging Face team for the 🤗 Datasets library
- The open-source NLP community

---
*Dataset created and published by Arabov Mullosharaf Kurbonovich using Hugging Face 🤗 Datasets*