shiima/vejin-Dataset-Normalization

Kurdish Books Dataset (Preprocessed)

Dataset Description

This dataset contains 18,565 Kurdish books with asosoft preprocessing applied to the content field. The dataset was created from an Excel file and includes book metadata along with preprocessed text content.

Languages

• Central Kurdish (ckb)
• Kurdish (ku)

Dataset Structure

The dataset contains the following columns:

• author
• book
• title
• url
• content

Data Processing

Text Preprocessing

The content column was preprocessed using the asosoft library with the following pipeline:

1. Text normalization (asosoft.Normalize) - Standardizes Kurdish characters and encoding
2. Punctuation normalization (asosoft.NormalizePunctuations) - Standardizes punctuation marks
3. Line trimming (asosoft.TrimLine) - Removes extra whitespace

This preprocessing ensures:

• Consistent Kurdish character encoding
• Standardized punctuation
• Removal of extra whitespace
• Improved text quality for NLP tasks

Example Usage

from datasets import load_dataset

# Load the dataset
dataset = load_dataset("shiima/vejin-Dataset-Normalization")

# View a sample
sample = dataset['train'][0]
print(f"Author: {sample['author']}")
print(f"Book: {sample['book']}")
print(f"Title: {sample['title']}")
print(f"URL: {sample['url']}")
print(f"Content: {sample['content'][:200]}...")

# Iterate through samples
for sample in dataset['train']:
    author = sample['author']
    content = sample['content']
    print(f"{author}: {content[:100]}...")
    break

Use Cases

This dataset is suitable for:

• Kurdish language model training
• Kurdish book analysis and understanding
• Kurdish text generation
• Kurdish NLP research
• Information extraction from Kurdish books
• Kurdish literature studies
• Document understanding tasks

Statistics

• Total Samples: 18,565
• Languages: Kurdish (Central Kurdish, Kurmanji)
• Script: Arabic script
• Format: Structured book data with metadata
• Columns: 5 (author, book, title, url, content)
• Preprocessing: asosoft normalization applied to content

Data Fields

• author: Name of the book's author
• book: Book name or identifier
• title: Title of the text/chapter
• url: Source URL or reference
• content: Main text content (preprocessed with asosoft)

Quality

The content has been preprocessed to ensure:

• Consistent character encoding across all books
• Standardized punctuation
• Clean text suitable for NLP applications
• Removed formatting artifacts

Source

This dataset was created from an Excel file containing Kurdish book data. The content field was preprocessed using asosoft for improved text quality.

Citation

If you use this dataset, please acknowledge:

• The asosoft preprocessing library: asosoft
• The original book sources (see URL field)

License

Apache 2.0 (or specify your license)

Acknowledgments

• Preprocessing library: asosoft by Kurdish NLP
• Original book authors and sources

Notes

• All content has been preprocessed for consistency
• The dataset maintains original metadata (author, book, title, URL)
• Suitable for various Kurdish NLP tasks
• Ready for language model training and fine-tuning

Contact

For questions or issues, please open an issue on the dataset repository.