metadata

dataset_info:
  features:
    - name: text
      dtype: string
  splits:
    - name: train
      num_bytes: 45957987986
      num_examples: 16896817
  download_size: 21312867175
  dataset_size: 45957987986
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
task_categories:
  - text-generation
language:
  - fa
pretty_name: 'HmBlogs: A big general Persian corpus'
size_categories:
  - 10M<n<100M

HmBlogs: A big general Persian corpus

HmBlogs is a general Persian corpus collected from nearly 20 million blog posts over a period of 15 years containig 6.8 billion tokens. This version is the preprocessed version of the dataset prepared by the original authors and converted to proper format to integrate with 🤗Datasets. In order to access the raw versions visit the official link at http://nlplab.sbu.ac.ir/hmBlogs-v3 .

Paper: https://arxiv.org/abs/2111.02362
Authors: Hamzeh Motahari Khansari, Mehrnoush Shamsfard
Original Link: http://nlplab.sbu.ac.ir/hmBlogs-v3/

Usage

This dataset can be used for masked/causal language modeling. You can easily load this dataset like below:

from datasets import load_dataset

# Load the whole dataset
dataset = load_dataset("arxyzan/hmblogs-clean", split="train")

# Load a portion by %
dataset = load_dataset("arxyzan/hmblogs-clean", split="train[:50%]")

# Load a custom shard
dataset = load_dataset("arxyzan/hmblogs-clean", data_files=["data/train-00000-of-00046.parquet", "data/train-00001-of-00046.parquet"])

Citation

@article{DBLP:journals/corr/abs-2111-02362,
  author       = {Hamzeh Motahari Khansari and
                  Mehrnoush Shamsfard},
  title        = {HmBlogs: {A} big general Persian corpus},
  journal      = {CoRR},
  volume       = {abs/2111.02362},
  year         = {2021},
  url          = {https://arxiv.org/abs/2111.02362},
  eprinttype    = {arXiv},
  eprint       = {2111.02362},
  timestamp    = {Fri, 05 Nov 2021 15:25:54 +0100},
  biburl       = {https://dblp.org/rec/journals/corr/abs-2111-02362.bib},
  bibsource    = {dblp computer science bibliography, https://dblp.org}
}