amitca71/parsoff_DeepSentiPers_ParsiNLU

Persian Human Text Dataset (Subset of Parsoff, DeepSentiPers, ParsiNLU)

This dataset contains human-authored Persian (Farsi) text collected as a subset and aggregation of several well-known Persian NLP datasets.

The purpose of this dataset is to provide a clean, consolidated reference corpus of human-written Persian text, primarily for comparison with synthetic or LLM-generated datasets.

---

Dataset Origin

This dataset is composed exclusively of samples drawn from the following original datasets:

1. Parsoff
2. DeepSentiPers
3. ParsiNLU

No synthetic or machine-generated text is included. The dataset does not introduce new annotations and does not alter the original content beyond filtering, consolidation, and formatting.

---

Dataset Structure

Columns

• text: Human-authored Persian text (reviews, social media posts, handwriting transcriptions).
• label: Task-specific label inherited from the original dataset.
• source_dataset: Origin of the sample:
  • parsoff
  • DeepSentiPers
  • ParsiNLU

---

Intended Use

This dataset is intended for Non-Commercial Research Only, aligning with the licenses of the source data:

• Academic research in Persian NLP.
• Benchmarking against synthetic / LLM-generated text.
• Sentiment analysis and NLU experiments.
• Stylometric and linguistic analysis.

---

Personal and Sensitive Information

Disclaimer: This dataset contains text from social media, public reviews, and handwriting samples. While the original datasets may have undergone anonymization by their respective authors, users should be aware that the text may still contain Personally Identifiable Information (PII).

Users are responsible for ensuring their use of this data complies with applicable privacy laws (e.g., GDPR) and ethical guidelines regarding data privacy.

---

Dataset Takedown Policy

This dataset is a derivative work aggregated for research purposes. We make no claim of ownership over the original data.

If you are a rights holder of any of the original datasets or an individual whose data is included, and you wish for it to be removed, please open a Discussion in the Community tab of this repository, and we will remove the relevant data immediately.

---

License

This dataset is an aggregation of subsets from multiple sources and does not introduce a new unified license.

Each sample in the dataset remains subject to the original license of its source dataset:

• ParsiNLU: CC BY-NC-SA 4.0
• DeepSentiPers: Research Use Only / Non-Commercial
• Parsoff: Research Use Only

Users must comply with the license terms of the original dataset corresponding to each sample, as indicated by the source_dataset field.

No relicensing is implied.

---

Credits & Citations

You must cite the original authors when using this dataset. This dataset is merely a derived view; the intellectual credit belongs to the following works:

1. ParsiNLU

Khashabi, D., et al. (2021).

• Paper: Read on ArXiv
• Code: GitHub Repository

@article{khashabi2020parsinlu,
  title={ParsiNLU: A Suite of Language Understanding Datasets for Persian},
  author={Khashabi, Daniel and Cohan, Arman and Shakeri, Siamak and Hosseini, Pedram and Pezeshkpour, Pouya and Alikhani, Malihe and Aminnaseri, Moin and Bitaab, Marjan and Sadeqi, Faeze and Habibian, Ahmad and others},
  journal={arXiv preprint arXiv:2012.06154},
  year={2020}
}

@article{sharami2020deepsentipers,
  title={DeepSentiPers: Novel Deep Learning Models Trained Over Proposed Augmented Persian Sentiment Corpus},
  author={Sharami, Javad PourMostafa Roshan and Sarabestani, Parsa Abbasi and Mirroshandel, Seyed Abolghasem},
  journal={arXiv preprint arXiv:2004.05328},
  year={2020}
}

@article{ebrahimi2022parsoff,
  title={Parsoff: A specific dataset for offline handwritten Persian recognition},
  author={Ebrahimi, M. and Torkaman, A.},
  journal={Journal of AI and Data Mining},
  volume={10},
  number={1},
  pages={111--120},
  year={2022},
  publisher={Shahrood University of Technology}
}