---
language:
- tg
license: apache-2.0
tags:
- morphological-corpus
- tajik
- grammar
- linguistic
---

# Tajik Unified Morphological Corpus

**Author:** Arabov Mullosharaf Kurbonovich

Morphological corpus of the Tajik language, combining data from [tajik-corpus.org](https://tajik-corpus.org).  
Contains wordforms, lemmas, grammatical tags, and frequency information.

## 📖 Description
The dataset includes **200,144** entries (wordforms with their morphological analyses) with:

- wordform (word)
- lemma (base form)
- grammar tag (grammar) - combines POS and grammatical features
- part of speech (pos)
- grammatical features (features)
- gloss (short meaning)
- English translation (trans_en)
- Russian translation (trans_ru)
- raw grammar tag (gramm_raw)
- frequency (freq) - number of occurrences in the corpus

## 📊 Statistics

| Metric | Value |
|--------|-------|
| Total entries | 200,144 |
| Unique wordforms | 135,595 |
| Unique lemmas | 20,618 |
| Unique grammar tags | 10,287 |
| Total frequency sum | 200,144 |
| Mean frequency per entry | 1.0 |
| Median frequency | 1.0 |

### Part of Speech Distribution
| Part of Speech   | Count   |
|:-----------------|:--------|
| N                | 126,907 |
| ADJ              | 44,995  |
| V                | 21,501  |
| ADV              | 3,975   |
| PRON             | 886     |
| NUM              | 774     |
| CONJ             | 383     |
| INTJ             | 300     |
| PRTCL            | 285     |
| PREP             | 92      |
| part             | 31      |
| POST             | 11      |
| INTZ             | 3       |
| PROP             | 1       |

### Most Frequent Words
| Word        |   Frequency |
|:------------|------------:|
| бозоргонам  |          12 |
| бозоргоне   |          12 |
| бозоргонеро |          12 |
| бозӣ        |          12 |
| баротӣ      |          11 |
| баротам     |          10 |
| баҳодурӣ    |          10 |
| бозорам     |          10 |
| бозоре      |          10 |
| бозореро    |          10 |

### Most Frequent Lemmas
| Lemma   |   Frequency |
|:--------|------------:|
| сар     |         276 |
| бар     |         267 |
| овардан |         226 |
| ёфтан   |         223 |
| кардан  |         215 |
| гуфтан  |         202 |
| боз     |         201 |
| додан   |         194 |
| кор     |         194 |
| омадан  |         193 |

### Most Frequent Grammar Tags
| Grammar Tag   | Count   |
|:--------------|:--------|
| N,sg          | 9,077   |
| N,sg,ezf      | 7,579   |
| N,and         | 5,502   |
| N,sg,and      | 5,502   |
| ADJ,sg        | 4,589   |
| N,sg,obj.def  | 3,441   |
| ADJ,sg,ezf    | 3,221   |
| N,sg,indef    | 2,995   |
| N,sg,rel      | 2,995   |
| N,pl,ezf      | 2,783   |

### Word Length Statistics (in characters)
| Statistic   |   Characters |
|:------------|-------------:|
| Mean        |            8 |
| Median      |            8 |
| Minimum     |            1 |
| Maximum     |           21 |

### Number of Grammatical Features per Wordform
| Statistic   |   Number of Tags |
|:------------|-----------------:|
| Mean        |              4.6 |
| Median      |              4   |
| Minimum     |              1   |
| Maximum     |             15   |

## 🚀 Usage Example

```python
from datasets import load_dataset

# Load the dataset
dataset = load_dataset("TajikNLPWorld/TajikUnifiedCorpus")
train = dataset["train"]

# Filter nouns
nouns = train.filter(lambda x: x["pos"] == "N")

# Filter verbs in past tense
verbs_past = train.filter(lambda x: "past" in x["features"] if x["features"] else False)

# Top 10 most frequent words
import pandas as pd
df = train.to_pandas()
top_words = df.groupby("word")["freq"].sum().nlargest(10)
print(top_words)

# Iterate through records
for record in train.select(range(5)):
    print(f"{record['word']} -> {record['lemma']} ({record['grammar']})")
```

## 🔬 Possible Applications

* Morphological analysis of Tajik language
* Part-of-speech tagging training
* Linguistic research
* Dictionary and educational material creation
* Grammatical pattern extraction
* Machine translation systems

## 📜 Citation

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

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

## 📜 License

Apache 2.0