Bert Base Parsbert Ner Uncased

🚀 ParsBERT: Transformer-based Model for Persian Language Understanding

ParsBERT is a monolingual language model based on Google’s BERT architecture, with the same configurations as BERT-Base. It aims to enhance Persian language understanding.

The paper presenting ParsBERT can be found at arXiv:2005.12515.

All the models (downstream tasks) are uncased and trained with whole word masking. (coming soon stay tuned)

✨ Features

Persian NER [ARMAN, PEYMA, ARMAN+PEYMA]

This task focuses on extracting named entities in the text, such as names, and labeling them with appropriate NER classes like locations, organizations, etc. The datasets used for this task contain sentences marked in the IOB format. In this format, tokens not part of an entity are tagged as ”O”, the ”B” tag corresponds to the first word of an object, and the ”I” tag corresponds to the rest of the terms of the same entity. Both ”B” and ”I” tags are followed by a hyphen (or underscore), then the entity category. So, the NER task is a multi - class token classification problem that labels tokens when fed raw text. There are two primary datasets used in Persian NER, ARMAN and PEYMA. In ParsBERT, we prepared NER for both datasets and their combination.

PEYMA

The PEYMA dataset includes 7,145 sentences with a total of 302,530 tokens, of which 41,148 tokens are tagged with seven different classes:

1. Organization
2. Money
3. Location
4. Date
5. Time
6. Person
7. Percent

Download: You can download the dataset from [here](http://nsurl.org/tasks/task - 7 - named - entity - recognition - ner - for - farsi/)

ARMAN

The ARMAN dataset holds 7,682 sentences with 250,015 tokens tagged over six different classes:

1. Organization
2. Location
3. Facility
4. Event
5. Product
6. Person

Download: You can download the dataset from here

📊 Results

The following table summarizes the F1 score obtained by ParsBERT compared to other models and architectures:

💻 Usage Examples

Basic Usage

📚 Documentation

Cite

Please cite the following paper in your publication if you are using ParsBERT in your research:

@article{ParsBERT,
    title={ParsBERT: Transformer-based Model for Persian Language Understanding},
    author={Mehrdad Farahani, Mohammad Gharachorloo, Marzieh Farahani, Mohammad Manthouri},
    journal={ArXiv},
    year={2020},
    volume={abs/2005.12515}
}

Acknowledgments

We express our gratitude to the Tensorflow Research Cloud (TFRC) program for providing the necessary computation resources. We also thank the Hooshvare Research Group for facilitating dataset gathering and scraping online text resources.

Contributors

• Mehrdad Farahani: Linkedin, Twitter, Github
• Mohammad Gharachorloo: [Linkedin](https://www.linkedin.com/in/mohammad - gharachorloo/), Twitter, Github
• Marzieh Farahani: Linkedin, Twitter, Github
• Mohammad Manthouri: [Linkedin](https://www.linkedin.com/in/mohammad - manthouri - aka - mansouri - 07030766/), Twitter, Github
• Hooshvare Team: Official Website, Linkedin, Twitter, Github, Instagram

A special thanks to Sara Tabrizi for her fantastic poster design. Follow her on: [Linkedin](https://www.linkedin.com/in/sara - tabrizi - 64548b79/), Behance, Instagram

Releases

Release v0.1 (May 29, 2019)

This is the first version of our ParsBERT NER!

📄 License

This project is licensed under the Apache - 2.0 license.