aictsharif/perisan_texts_with_typos

Persian Typo Correction Dataset

Dataset Summary

The Persian Typo Correction Dataset is a parallel corpus consisting of Persian (Farsi) sentences with common typographical, spelling, and orthographic errors alongside their corrected versions. This resource is designed for training and evaluating machine learning models for automatic text correction, normalization, and spell checking in Persian. It is suitable for both academic research and real-world NLP applications, particularly those focusing on Persian language processing.

Data Structure

Each sample in the dataset consists of:

• with_typos: A Persian sentence containing natural or simulated typographical and spelling errors.
• fixed: The same sentence with all errors corrected.

The dataset is available in CSV format with two columns:

• with_typos: string, sentence with typos
• fixed: string, corrected sentence

Example

with_typos	fixed
وزیر امور خارجه در سخنانی اظعار داشت که ایران به دنبال تقوییت روابط با همسایگان است.	وزیر امور خارجه در سخنانی اظهار داشت که ایران به دنبال تقویت روابط با همسایگان است.
نمایشگاه بین المللی کتاب تهران، فرصتی مناسب برای اشنایی با اخرین اثار منتشره در حوزه کتاب و نشر است.	نمایشگاه بین‌المللی کتاب تهران، فرصتی مناسب برای آشنایی با آخرین آثار منتشر شده در حوزه کتاب و نشر است.
نمایشگاه بین المللی کتاب تهران با حضور ناشران داخلی و خارجی در شهر آفتاب برگزار میشود.	نمایشگاه بین المللی کتاب تهران با حضور ناشران داخلی و خارجی در شهر آفتاب برگزار می‌شود.

Dataset Applications

This dataset can be used for:

• Spell and Grammar Correction: Training sequence-to-sequence models (e.g., T5, BART, mBART) to automatically correct spelling and grammar mistakes in Persian text.
• Data Cleaning Tools: Building preprocessing pipelines for Persian text normalization.
• NLP Research: Benchmarking and evaluating typo correction algorithms for low-resource languages.
• Text Input Assistance: Enhancing user experience in chatbots, virtual keyboards, and writing assistants for Persian speakers.

Intended Use

This dataset is intended for:

• Researchers developing Persian NLP models for error correction.
• Developers aiming to improve Persian language text input on websites, apps, or messaging services.
• Language technology companies interested in Persian language tools.

Dataset Format

• Files: train.csv, test.csv
• Encoding: UTF-8
• Language: Persian (Farsi)

Citation

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

@misc{persian_typo_correction_2024,
  title        = {Persian Typo Correction Dataset},
  author       = {AICT Sharif},
  year         = {2025},
  url          = {https://huggingface.co/datasets/aict-sharif-edu/perisan_texts_with_typos},
  note         = {A parallel corpus for typo correction in Persian.}
}

Contributors

• Masoud Kaviani
• Soheil Tehranipour