---
license: cc-by-nc-sa-4.0
dataset_info:
- config_name: anl-news
  features:
  - name: text
    dtype: string
  splits:
  - name: train
    num_bytes: 1500707584
    num_examples: 236443
  download_size: 773593491
  dataset_size: 1500707584
- config_name: azwiki
  features:
  - name: id
    dtype: int64
  - name: text
    dtype: string
  - name: title
    dtype: string
  splits:
  - name: train
    num_bytes: 360206818
    num_examples: 129433
  download_size: 204669909
  dataset_size: 360206818
- config_name: bhos
  features:
  - name: title
    dtype: string
  - name: text
    dtype: string
  - name: id
    dtype: int64
  splits:
  - name: train
    num_bytes: 736156688
    num_examples: 488390
  download_size: 417517945
  dataset_size: 736156688
- config_name: elite-blogs
  features:
  - name: id
    dtype: int64
  - name: source
    dtype: string
  - name: text
    dtype: string
  splits:
  - name: train
    num_bytes: 7625261
    num_examples: 755
  download_size: 4031201
  dataset_size: 7625261
- config_name: elite-books
  features:
  - name: text
    dtype: string
  - name: id
    dtype: int64
  splits:
  - name: train
    num_bytes: 38894982
    num_examples: 104
  download_size: 22016093
  dataset_size: 38894982
- config_name: eqanun
  features:
  - name: text
    dtype: string
  - name: title
    dtype: string
  - name: id
    dtype: int64
  splits:
  - name: train
    num_bytes: 404638424
    num_examples: 53656
  download_size: 149151917
  dataset_size: 404638424
- config_name: mediocore-books
  features:
  - name: ID
    dtype: string
  - name: ' Metadata'
    dtype: string
  - name: text
    dtype: string
  splits:
  - name: train
    num_bytes: 2908183660
    num_examples: 7807263
  download_size: 695603782
  dataset_size: 2908183660
- config_name: translated-enwiki
  features:
  - name: text
    dtype: string
  splits:
  - name: train
    num_bytes: 1629190007
    num_examples: 280465
  download_size: 919526548
  dataset_size: 1629190007
configs:
- config_name: anl-news
  data_files:
  - split: train
    path: anl-news/train-*
- config_name: azwiki
  data_files:
  - split: train
    path: azwiki/train-*
- config_name: bhos
  data_files:
  - split: train
    path: bhos/train-*
- config_name: elite-blogs
  data_files:
  - split: train
    path: elite-blogs/train-*
- config_name: elite-books
  data_files:
  - split: train
    path: elite-books/train-*
- config_name: eqanun
  data_files:
  - split: train
    path: eqanun/train-*
- config_name: mediocore-books
  data_files:
  - split: train
    path: mediocore-books/train-*
- config_name: translated-enwiki
  data_files:
  - split: train
    path: translated-enwiki/train-*
task_categories:
- fill-mask
language:
- az
size_categories:
- 1M<n<10M
---

If you use this dataset, please cite us:
```bib
@inproceedings{isbarov-etal-2024-open,
    title = "Open foundation models for {A}zerbaijani language",
    author = "Isbarov, Jafar  and
      Huseynova, Kavsar  and
      Mammadov, Elvin  and
      Hajili, Mammad and
      Ataman, Duygu",
    editor = {Ataman, Duygu  and
      Derin, Mehmet Oguz  and
      Ivanova, Sardana  and
      K{\"o}ksal, Abdullatif  and
      S{\"a}lev{\"a}, Jonne  and
      Zeyrek, Deniz},
    booktitle = "Proceedings of the First Workshop on Natural Language Processing for Turkic Languages (SIGTURK 2024)",
    month = aug,
    year = "2024",
    address = "Bangkok, Thailand and Online",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2024.sigturk-1.2",
    pages = "18--28",
    abstract = "The emergence of multilingual large language models has enabled the development of language understanding and generation systems in Azerbaijani. However, most of the production-grade systems rely on cloud solutions, such as GPT-4. While there have been several attempts to develop open foundation models for Azerbaijani, these works have not found their way into common use due to a lack of systemic benchmarking. This paper encompasses several lines of work that promote open-source foundation models for Azerbaijani. We introduce (1) a large text corpus for Azerbaijani, (2) a family of encoder-only language models trained on this dataset, (3) labeled datasets for evaluating these models, and (4) extensive evaluation that covers all major open-source models with Azerbaijani support.",
}
```