shiima/kurdish-unified-corpus

Unified Kurdish Corpus

Dataset Description

This dataset aggregates 758,166 Kurdish text samples from multiple high-quality sources. All text has been preprocessed using asosoft.

Languages

• Central Kurdish (ckb)
• Kurdish (ku)

Dataset Structure

Columns

• text: Preprocessed text content (asosoft applied)
• base_dataset: Source dataset name
• url: Source URL (NULL if not available)
• word_count: Number of words (space-separated)
• character_count: Number of characters
• source: Original source (URL, book name, or dataset name)

Source Datasets

1. FineWeb-2 - HuggingFaceFW/fineweb-2 (ckb_Arab)
2. FinePDFs - HuggingFaceFW/finepdfs (ckb_Arab)
3. Vejin Books - shiima/vejin-Dataset-Normalization-with-textColumn
4. CulturaX - uonlp/CulturaX (ckb)

Statistics

Overall

• Total Samples: 758,166
• Total Words: 447,612,687
• Total Characters: 2,807,525,881
• Average Words: 590.39
• Average Characters: 3703.05
• Median Words: 183
• Samples with URL: 758,166 (100.00%)

By Dataset

Dataset	Samples	Avg Words	Avg Characters
FineWeb-2	554993	388.738	2524.92
FinePDFs	12574	12058.1	70500.3
Vejin Books	18565	252.893	1389.12
CulturaX	172034	439.166	2871.26

Data Processing

All text was preprocessed using asosoft:

1. Text normalization (asosoft.Normalize)
2. Punctuation normalization (asosoft.NormalizePunctuations)
3. Line trimming (asosoft.TrimLine)

Example Usage

from datasets import load_dataset

dataset = load_dataset("shiima/kurdish-unified-corpus")

# View sample
sample = dataset['train'][0]
print(f"Text: {sample['text'][:100]}...")
print(f"Dataset: {sample['base_dataset']}")
print(f"Source: {sample['source']}")
print(f"Words: {sample['word_count']}")

# Filter by dataset
fineweb = dataset['train'].filter(lambda x: x['base_dataset'] == 'FineWeb-2')

# Filter by word count
long_texts = dataset['train'].filter(lambda x: x['word_count'] > 100)

License

Apache 2.0

Acknowledgments

• Source datasets creators
• asosoft by Kurdish NLP