---
language:
- tg
license: apache-2.0
tags:
- morphological-corpus
- tajik
- grammar
- linguistic
pretty_name: Tajik Morphological Corpus
---

# Tajik Morphological Corpus

A morphological corpus of the Tajik language containing word forms, lemmas, grammatical tags, and frequency information.

## 📖 Description
The dataset includes **3,609** unique word forms (after deduplication) with:

- word form (`word`)
- lemma (`lemma`)
- grammatical tags (`grammar`), separated by commas and a vertical bar for alternative analyses
- frequency (`freq`) – number of occurrences in the original corpus

## 📊 Statistics

| Metric | Value |
|--------|-------|
| Total entries | 3,609 |
| Unique word forms | 3,460 |
| Unique lemmas | 2,126 |
| Unique grammar tags | 1,988 |
| Total frequency | 2,820,977 |
| Average frequency per entry | 781.7 |
| Median frequency | 3.0 |

### Most frequent words
| Word   | Frequency |
|:-------|----------:|
| ин     | 714,184  |
| барои  | 315,952  |
| аст    | 303,700  |
| то     | 191,645  |
| онҳо   | 144,608  |
| бар    | 106,422  |
| ҳамин  | 79,441   |
| ту     | 77,709   |
| нест   | 59,390   |
| пас    | 55,807   |

### Most frequent lemmas
| Lemma      | Frequency |
|:-----------|----------:|
| ин         | 724,582  |
| аст        | 322,903  |
| барои      | 319,891  |
| та/то      | 191,645  |
| он         | 188,309  |
| бар/бурдан | 106,422  |
| ҳам/ҳамин  | 80,561   |
| ту         | 77,796   |
| не/нест    | 59,391   |
| пас        | 55,823   |

### Most frequent grammar tags
| Grammar tag                                                                               | Count |
|:------------------------------------------------------------------------------------------|------:|
| cnject2, indir, prs, 2, sg, V | part, part.mod.prs, part.mod, V                            | 43   |
| hab.part.pst, part, sbjv.hab, V                                                           | 42   |
| inf, prs, 2, sg, V, cop, cop.encl | part, part.fut, V                                          | 40   |
| pst, fut, 3, sg, V                                                                        | 30   |
| pass.part, part.pst, pass.part.pst, part, V                                               | 28   |
| part, part.prs, V                                                                         | 27   |
| indir, imp, 2, neg, neg2, sg, V                                                           | 27   |
| part.pst, part, pluprf, V                                                                 | 26   |
| part.pst, part, pst.prog, prog, V                                                         | 26   |
| part.pst, part, sbjv.pst, V                                                               | 24   |

### Word length (characters)
| Statistic | Characters |
|:----------|-----------:|
| Mean      | 8.4       |
| Median    | 8         |
| Minimum   | 2         |
| Maximum   | 18        |

### Number of grammar tags per word form
| Statistic | Number of tags |
|:----------|---------------:|
| Mean      | 8.1           |
| Median    | 6             |
| Minimum   | 1             |
| Maximum   | 59            |

## 🚀 Usage example

```python
from datasets import load_dataset

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

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

# Find all nouns (tag contains "N")
nouns = train.filter(lambda x: "N" in x["grammar"])

# Top 10 most frequent words
top_words = train.to_pandas().groupby("word")["freq"].sum().nlargest(10)

# Iterate through records
for record in train:
    print(record["word"], record["lemma"], record["grammar"], record["freq"])
```

## 🔬 Potential applications
- Morphological analysis of Tajik
- Training part‑of‑speech taggers
- Linguistic research on grammatical categories
- Creating dictionaries and teaching materials

## 📜 License
Apache 2.0

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

```bibtex
@dataset{tajik_morph_corpus_2026,
    title = {Tajik Morphological Corpus},
    author = {Arabov Mullosharaf Kurbonovich (TajikNLPWorld)},
    year = {2026},
    publisher = {Hugging Face},
    url = {https://huggingface.co/datasets/TajikNLPWorld/TajikMorphCorpus}
}