TajikNLPWorld/tg-fa-parallel-lexicon

TajPersParallelLexicalCorpus

A parallel lexical corpus for Tajik–Persian language pairs. Each record contains a Tajik word/phrase, its Persian equivalent, and the source author or work.

📊 Dataset Statistics

Overview

Metric	Value
Total entries	175,130
Unique Tajik words	174,298
Unique Persian words	173,550
Unique authors / sources	16

Data completeness

All fields are fully populated – no missing values.

Field	Empty records
Tajik (tajik)	0
Persian (persian)	0
Author (author)	0

Origin of the data

The corpus was built by merging three CSV files:

Source file	Entries	Percentage
words.csv	49,758	28.41%
poetry_parts.csv	101,380	57.89%
prose_parts.csv	23,992	13.70%

Top‑10 authors by number of entries

Author	Entries	Percentage
Dictionary	49,758	28.41%
Rumi	31,275	17.86%
Ferdousi	22,290	12.73%
Nezami	14,926	8.52%
parsianjoman	10,844	6.19%
BBC	10,700	6.11%
Asadi	7,797	4.45%
Saadi	7,593	4.34%
Bidel	4,443	2.54%
Hafez	4,018	2.29%

Note: "Dictionary" entries come from a bilingual wordlist; the remaining authors are classical and modern poets/prose writers (Rumi, Ferdousi, Nezami, Saadi, Hafez, etc.) and sources like BBC and parsianjoman.

📁 Data Format

Each record contains three fields:

Field	Type	Description
tajik	string	Word/phrase in Tajik (Cyrillic script)
persian	string	Corresponding word/phrase in Persian (Arabic script)
author	string	Source author, work, or dictionary name

🚀 Usage Example

from datasets import load_dataset

dataset = load_dataset("TajikNLPWorld/TajPersParallelLexicalCorpus")
data = dataset["train"]

# First record
print(data[0])

# Filter by author
saadi_entries = data.filter(lambda x: x["author"] == "Saadi")
print(f"Entries from Saadi: {len(saadi_entries)}")

# Get all unique authors
authors = set(data["author"])
print(f"All authors: {authors}")

Sample record

{
    "tajik": "модар",
    "persian": "مادر",
    "author": "Dictionary"
}

🔬 Potential Applications

• Tajik–Persian machine translation
• Cross‑lingual word embedding learning
• Lexicography and linguistic research
• Educational tools for Persian speakers learning Tajik (and vice versa)
• Literary studies and authorship analysis

📜 License

Apache 2.0

🤝 Citation

@dataset{tajpers_parallel_lexical_corpus_2026,
    title = {TajPersParallelLexicalCorpus},
    author = {TajikNLPWorld},
    year = {2026},
    publisher = {Hugging Face},
    url = {https://huggingface.co/datasets/TajikNLPWorld/TajPersParallelLexicalCorpus}
}

📬 Contact

For questions or contributions, please open an issue on the Hugging Face repository or contact the TajikNLPWorld team.

---

Automatically generated from CSV sources (words.csv, poetry_parts.csv, prose_parts.csv).