README.md

---
license: cc-by-nc-4.0
task_categories:
- fill-mask
- text-generation
language:
- am
- ar
- ay
- bm
- bbj
- bn
- bs
- bg
- ca
- cs
- ku
- da
- el
- en
- et
- ee
- fil
- fi
- fr
- fon
- gu
- guw
- ha
- he
- hi
- hu
- ig
- id
- it
- ja
- kk
- km
- ko
- lv
- ln
- lt
- lg
- luo
- mk
- mos
- my
- nl
- 'no'
- ne
- om
- or
- pa
- pcm
- fa
- pl
- pt
- mg
- ro
- rn
- ru
- sn
- so
- es
- sr
- sq
- sw
- sv
- ta
- tet
- ti
- th
- tn
- tr
- tw
- uk
- ur
- wo
- xh
- yo
- zh
- zu
- de
multilinguality:
- multilingual
pretty_name: PolyNews
size_categories:
- 1K<n<10K
source_datasets:
- masakhanews
- mafand
- wikinews
- wmt-news
- globalvoices
tags:
- news
- polynews
- mafand
- masakhanews
- wikinews
- globalvoices
- wmtnews
configs:
- config_name: amh_Ethi
  data_files:
  - split: train
    path: data/amh_Ethi/train.parquet.gzip
- config_name: arb_Arab
  data_files:
  - split: train
    path: data/arb_Arab/train.parquet.gzip
- config_name: ayr_Latn
  data_files:
  - split: train
    path: data/ayr_Latn/train.parquet.gzip
- config_name: bam_Latn
  data_files:
  - split: train
    path: data/bam_Latn/train.parquet.gzip
- config_name: bbj_Latn
  data_files:
  - split: train
    path: data/bbj_Latn/train.parquet.gzip
- config_name: ben_Beng
  data_files:
  - split: train
    path: data/ben_Beng/train.parquet.gzip
- config_name: bos_Latn
  data_files:
  - split: train
    path: data/bos_Latn/train.parquet.gzip
- config_name: bul_Cyrl
  data_files:
  - split: train
    path: data/bul_Cyrl/train.parquet.gzip
- config_name: cat_Latn
  data_files:
  - split: train
    path: data/cat_Latn/train.parquet.gzip
- config_name: ces_Latn
  data_files:
  - split: train
    path: data/ces_Latn/train.parquet.gzip
- config_name: ckb_Arab
  data_files:
  - split: train
    path: data/ckb_Arab/train.parquet.gzip
- config_name: dan_Latn
  data_files:
  - split: train
    path: data/dan_Latn/train.parquet.gzip
- config_name: deu_Latn
  data_files:
  - split: train
    path: data/deu_Latn/train.parquet.gzip
- config_name: ell_Grek
  data_files:
  - split: train
    path: data/ell_Grek/train.parquet.gzip
- config_name: eng_Latn
  data_files:
  - split: train
    path: data/eng_Latn/train.parquet.gzip
- config_name: est_Latn
  data_files:
  - split: train
    path: data/est_Latn/train.parquet.gzip
- config_name: ewe_Latn
  data_files:
  - split: train
    path: data/ewe_Latn/train.parquet.gzip
- config_name: fil_Latn
  data_files:
  - split: train
    path: data/fil_Latn/train.parquet.gzip
- config_name: fin_Latn
  data_files:
  - split: train
    path: data/fin_Latn/train.parquet.gzip
- config_name: fon_Latn
  data_files:
  - split: train
    path: data/fon_Latn/train.parquet.gzip
- config_name: fra_Latn
  data_files:
  - split: train
    path: data/fra_Latn/train.parquet.gzip
- config_name: guj_Gujr
  data_files:
  - split: train
    path: data/guj_Gujr/train.parquet.gzip
- config_name: guw_Latn
  data_files:
  - split: train
    path: data/guw_Latn/train.parquet.gzip
- config_name: hau_Latn
  data_files:
  - split: train
    path: data/hau_Latn/train.parquet.gzip
- config_name: heb_Hebr
  data_files:
  - split: train
    path: data/heb_Hebr/train.parquet.gzip
- config_name: hin_Deva
  data_files:
  - split: train
    path: data/hin_Deva/train.parquet.gzip
- config_name: hun_Latn
  data_files:
  - split: train
    path: data/hun_Latn/train.parquet.gzip
- config_name: ibo_Latn
  data_files:
  - split: train
    path: data/ibo_Latn/train.parquet.gzip
- config_name: ind_Latn
  data_files:
  - split: train
    path: data/ind_Latn/train.parquet.gzip
- config_name: ita_Latn
  data_files:
  - split: train
    path: data/ita_Latn/train.parquet.gzip
- config_name: jpn_Jpan
  data_files:
  - split: train
    path: data/jpn_Jpan/train.parquet.gzip
- config_name: kaz_Cyrl
  data_files:
  - split: train
    path: data/kaz_Cyrl/train.parquet.gzip
- config_name: khm_Khmr
  data_files:
  - split: train
    path: data/khm_Khmr/train.parquet.gzip
- config_name: kor_Hang
  data_files:
  - split: train
    path: data/kor_Hang/train.parquet.gzip
- config_name: lav_Latn
  data_files:
  - split: train
    path: data/lav_Latn/train.parquet.gzip
- config_name: lin_Latn
  data_files:
  - split: train
    path: data/lin_Latn/train.parquet.gzip
- config_name: lit_Latn
  data_files:
  - split: train
    path: data/lit_Latn/train.parquet.gzip
- config_name: lug_Latn
  data_files:
  - split: train
    path: data/lug_Latn/train.parquet.gzip
- config_name: luo_Latn
  data_files:
  - split: train
    path: data/luo_Latn/train.parquet.gzip
- config_name: mkd_Cyrl
  data_files:
  - split: train
    path: data/mkd_Cyrl/train.parquet.gzip
- config_name: mos_Latn
  data_files:
  - split: train
    path: data/mos_Latn/train.parquet.gzip
- config_name: mya_Mymr
  data_files:
  - split: train
    path: data/mya_Mymr/train.parquet.gzip
- config_name: nld_Latn
  data_files:
  - split: train
    path: data/nld_Latn/train.parquet.gzip
- config_name: nor_Latn
  data_files:
  - split: train
    path: data/nor_Latn/train.parquet.gzip
- config_name: npi_Deva
  data_files:
  - split: train
    path: data/npi_Deva/train.parquet.gzip
- config_name: orm_Latn
  data_files:
  - split: train
    path: data/orm_Latn/train.parquet.gzip
- config_name: ory_Orya
  data_files:
  - split: train
    path: data/ory_Orya/train.parquet.gzip
- config_name: pan_Guru
  data_files:
  - split: train
    path: data/pan_Guru/train.parquet.gzip
- config_name: pcm_Latn
  data_files:
  - split: train
    path: data/pcm_Latn/train.parquet.gzip
- config_name: pes_Arab
  data_files:
  - split: train
    path: data/pes_Arab/train.parquet.gzip
- config_name: plt_Latn
  data_files:
  - split: train
    path: data/plt_Latn/train.parquet.gzip
- config_name: pol_Latn
  data_files:
  - split: train
    path: data/pol_Latn/train.parquet.gzip
- config_name: por_Latn
  data_files:
  - split: train
    path: data/por_Latn/train.parquet.gzip
- config_name: ron_Latn
  data_files:
  - split: train
    path: data/ron_Latn/train.parquet.gzip
- config_name: run_Latn
  data_files:
  - split: train
    path: data/run_Latn/train.parquet.gzip
- config_name: rus_Cyrl
  data_files:
  - split: train
    path: data/rus_Cyrl/train.parquet.gzip
- config_name: sna_Latn
  data_files:
  - split: train
    path: data/sna_Latn/train.parquet.gzip
- config_name: som_Latn
  data_files:
  - split: train
    path: data/som_Latn/train.parquet.gzip
- config_name: spa_Latn
  data_files:
  - split: train
    path: data/spa_Latn/train.parquet.gzip
- config_name: sqi_Latn
  data_files:
  - split: train
    path: data/sqi_Latn/train.parquet.gzip
- config_name: srp_Cyrl
  data_files:
  - split: train
    path: data/srp_Cyrl/train.parquet.gzip
- config_name: srp_Latn
  data_files:
  - split: train
    path: data/srp_Latn/train.parquet.gzip
- config_name: swe_Latn
  data_files:
  - split: train
    path: data/swe_Latn/train.parquet.gzip
- config_name: swh_Latn
  data_files:
  - split: train
    path: data/swh_Latn/train.parquet.gzip
- config_name: tam_Taml
  data_files:
  - split: train
    path: data/tam_Taml/train.parquet.gzip
- config_name: tet_Latn
  data_files:
  - split: train
    path: data/tet_Latn/train.parquet.gzip
- config_name: tha_Thai
  data_files:
  - split: train
    path: data/tha_Thai/train.parquet.gzip
- config_name: tir_Ethi
  data_files:
  - split: train
    path: data/tir_Ethi/train.parquet.gzip
- config_name: tsn_Latn
  data_files:
  - split: train
    path: data/tsn_Latn/train.parquet.gzip
- config_name: tur_Latn
  data_files:
  - split: train
    path: data/tur_Latn/train.parquet.gzip
- config_name: twi_Latn
  data_files:
  - split: train
    path: data/twi_Latn/train.parquet.gzip
- config_name: ukr_Cyrl
  data_files:
  - split: train
    path: data/ukr_Cyrl/train.parquet.gzip
- config_name: urd_Arab
  data_files:
  - split: train
    path: data/urd_Arab/train.parquet.gzip
- config_name: wol_Latn
  data_files:
  - split: train
    path: data/wol_Latn/train.parquet.gzip
- config_name: xho_Latn
  data_files:
  - split: train
    path: data/xho_Latn/train.parquet.gzip
- config_name: yor_Latn
  data_files:
  - split: train
    path: data/yor_Latn/train.parquet.gzip
- config_name: zho_Hans
  data_files:
  - split: train
    path: data/zho_Hans/train.parquet.gzip
- config_name: zho_Hant
  data_files:
  - split: train
    path: data/zho_Hant/train.parquet.gzip
- config_name: zul_Latn
  data_files:
  - split: train
    path: data/zul_Latn/train.parquet.gzip
---
# Dataset Card for PolyNews

## Table of Contents
- [Table of Contents](#table-of-contents)
- [Dataset Description](#dataset-description)
  - [Dataset Summary](#dataset-summary)
  - [Uses](#uses)
  - [Languages](#languages)
- [Dataset Structure](#dataset-structure)
  - [Data Instances](#data-instances)
  - [Data Fields](#data-fields)
  - [Data Splits](#data-splits)
- [Dataset Creation](#dataset-creation)
  - [Curation Rationale](#curation-rationale)
  - [Source Data](#source-data)
  - [Data Collection and Processing](#data-collection-and-processing)
  - [Annotations](#annotations)
  - [Personal and Sensitive Information](#personal-and-sensitive-information)
- [Considerations for Using the Data](#considerations-for-using-the-data)
  - [Social Impact of Dataset](#social-impact-of-dataset)
  - [Discussion of Biases](#discussion-of-biases)
  - [Other Known Limitations](#other-known-limitations)
- [Additional Information](#additional-information)
  - [Licensing Information](#licensing-information)
  - [Citation Information](#citation-information)

## Dataset Description

- **Homepage:** https://huggingface.co/datasets/aiana94/polynews
- **Repository:** https://github.com/andreeaiana/nase 
- **Paper:**  [News Without Borders: Domain Adaptation of Multilingual Sentence Embeddings for Cross-lingual News Recommendation](https://arxiv.org/abs/2406.12634)
- **Point of Contact:** [Andreea Iana](https://andreeaiana.github.io/)
- **License:** [CC-BY-4.0-NC](https://creativecommons.org/licenses/by-nc/4.0/)

### Dataset Summary
PolyNews is a multilingual dataset containing news titles in 77 languages and 19 scripts.

### Uses

This dataset can be used for domain adaptation of language models, language modeling or text generation.

### Languages

There are 77 languages available:

|**Code**   |    **Language**             |   **Script**       |**\#Articles (K)**|
|:-----------|:---------------------------|:-------------------|:---------|
| amh\_Ethi |   Amharic                 | Ethiopic          | 0.551   |
| arb\_Arab |   Modern Standard Arabic  | Arabic            | 10.882  |
| ayr\_Latn |   Central Aymara          | Latin             | 12.878  |
| bam\_Latn |   Bambara                 | Latin             | 2.916   |
| bbj\_Latn |   Ghomálá’                | Latin             | 1.737   |
| ben\_Beng |   Bengali                 | Bengali           | 2.268   |
| bos\_Latn |   Bosnian                 | Latin             | 0.298   |
| bul\_Cyrl |   Bulgarian               | Cyrillic          | 1.791   |
| cat\_Latn |   Catalan                 | Latin             | 30.410  |
| ces\_Latn |   Czech                   | Latin             | 58.382  |
| ckb\_Arab |   Central Kurdish         | Arabic            | 0.014   |
| dan\_Latn |   Danish                  | Latin             | 9.456   |
| deu\_Latn |   German                  | Latin             | 145.484 |
| ell\_Grek |   Greek                   | Greek             | 50.176  |
| eng\_Latn |   English                 | Latin             | 981.430 |
| est\_Latn |   Estonian                | Latin             | 3.942   |
| ewe\_Latn |   Éwé                     | Latin             | 2.003   |
| fil\_Latn |   Filipino                | Latin             | 3.3132  |
| fin\_Latn |   Finnish                 | Latin             | 19.602  |
| fon\_Latn |   Fon                     | Latin             | 2.610   |
| fra\_Latn |   French                  | Latin             | 481.117 |
| guj\_Gujr |   Gujarati                | Gujarati          | 0.690   |
| guw\_Latn |   Gun                     | Latin             | 1.068   |
| hau\_Latn |   Hausa                   | Latin             | 7.898   |
| heb\_Hebr |   Hebrew                  | Hebrew            | 0.355   |
| hin\_Deva |   Hindi                   | Devanagari        | 0.707   |
| hun\_Latn |   Hungarian               | Latin             | 22.219  |
| ibo\_Latn |   Igbo                    | Latin             | 7.709   |
| ind\_Latn |   Indonesian              | Latin             | 17.749  |
| ita\_Latn |   Italian                 | Latin             | 163.396 |
| jpn\_Jpan |   Japanese                | Japanese          | 20.778  |
| kaz\_Cyrl |   Kazakh                  | Cyrillic          | 0.763   |
| khm\_Khmr |   Khmer                   | Khmer             | 0.227   |
| kor\_Hang |   Korean                  | Hangul            | 3.527   |
| lav\_Latn |   Latvian                 | Latin             | 3.971   |
| lin\_Latn |   Lingala                 | Latin             | 0.602   |
| lit\_Latn |   Lithuanian              | Latin             | 3.948   |
| lug\_Latn |   Ganda                   | Latin             | 4.769   |
| luo\_Latn |   Luo                     | Latin             | 4.250   |
| mkd\_Cyrl |   Macedonian              | Cyrillic          | 10.537  |
| mos\_Latn |   Mossi                   | Latin             | 2.458   |
| mya\_Mymr |   Burmese                 | Myanmar           | 0.583   |
| nld\_Latn |   Dutch                   | Latin             | 53.184  |
| nor\_Latn |   Norwegian               | Latin             | 0.529   |
| npi\_Deva |   Nepali                  | Devanagari        | 0.220   |
| orm\_Latn |   Oromo                   | Latin             | 1.124   |
| ory\_Orya |   Odia                    | Oriya             | 0.038   |
| pan\_Guru |   Eastern Panjabi         | Gurmukhi          | 0.336   |
| pcm\_Latn |   Nigerian Pidgin         | Latin             | 5.742   |
| pes\_Arab |   Western Persian         | Arabic            | 1.431   |
| plt\_Latn |   Malagasy                | Latin             | 393.767 |
| pol\_Latn |   Polish                  | Latin             | 80.960  |
| por\_Latn |   Portuguese              | Latin             | 156.039 |
| ron\_Latn |   Romanian                | Latin             | 10.472  |
| run\_Latn |   Rundi                   | Latin             | 1.113   |
| rus\_Cyrl |   Russian                 | Cyrillic          | 143.283 |
| sna\_Latn |   Shona                   | Latin             | 1.128   |
| som\_Latn |   Somali                  | Latin             | 1.019   |
| spa\_Latn |   Spanish                 | Latin             | 681.121 |
| sqi\_Latn |   Albanian                | Latin             | 7.274   |
| srp\_Cyrl |   Serbian                 | Cyrillic          | 1.056   |
| srp\_Latn |   Serbian                 | Latin             | 58.012  |
| swe\_Latn |   Swedish                 | Latin             | 12.323  |
| swh\_Latn |   Swahili                 | Latin             | 47.337  |
| tam\_Taml |   Tamil                   | Tamil             | 0.358   |
| tet\_Latn |   Tetun                   | Latin             | 0.626   |
| tha\_Thai |   Thai                    | Thai              | 0.091   |
| tir\_Ethi |   Tigrinya                | Ethiopic          | 0.079   |
| tsn\_Latn |   Tswana                  | Latin             | 2.075   |
| tur\_Latn |   Turkish                 | Latin             | 19.793  |
| twi\_Latn |   Twi                     | Latin             | 3.012   |
| ukr\_Cyrl |   Ukrainian               | Cyrillic          | 0.292   |
| urd\_Arab |   Urdu                    | Arabic            | 0.804   |
| wol\_Latn |   Wolof                   | Latin             | 3.344   |
| xho\_Latn |   Xhosa                   | Latin             | 0.709   |
| yor\_Latn |   Yorùbá                  | Latin             | 8.011   |
| zho\_Hans |   Chinese                 | Han (Simplified)  | 59.771  |
| zho\_Hant |   Chinese                 | Han (Traditional) | 54.561  |
| zul\_Latn |   Zulu                    | Latin             | 3.376   |


## Dataset Structure

### Data Instances
```
>>> from datasets import load_dataset
>>> data = load_dataset('aiana94/polynews', 'ron_Latn')

# Please, specify the language code,

# A data point example is below:

{
"text": "Un public numeros. Este uimitor succesul după doar trei ediții . ",
"provenance": "globalvoices"
}

```

### Data Fields

- text (string): news text
- provenance (string) : source dataset for the news example

### Data Splits

For all languages, there is only the `train` split.


## Dataset Creation

### Curation Rationale

Multiple multilingual, human-translated, datasets containing news texts have been released in recent years. 
However, these datasets are stored in different formats and various websites, and many contain numerous near duplicates. 
With PolyNews, we aim to provide an easily-accessible, unified and deduplicated dataset that combines these disparate data sources. 
It can be used for domain adaptation of language models, language modeling or text generation in both high-resource and low-resource languages.

### Source Data

The source data consists of five multilingual news datasets.

- [Wikinews](https://www.wikinews.org/) (latest dump available in May 2024)
- [GlobalVoices](https://opus.nlpl.eu/GlobalVoices/corpus/version/GlobalVoices) (v2018q4)
- [WMT-News](https://opus.nlpl.eu/WMT-News/corpus/version/WMT-News) (v2019)
- [MasakhaNews](https://huggingface.co/datasets/masakhane/masakhanews) (`train` split)
- [MAFAND](https://huggingface.co/datasets/masakhane/mafand) (`train` split)

#### Data Collection and Processing

We processed the data using a **working script** which covers the entire processing pipeline. It can be found [here](https://github.com/andreeaiana/nase/blob/main/scripts/construct_polynews.sh).

The data processing pipeline consists of:
1. Downloading the WMT-News and GlobalVoices News from OPUS.
2. Downloading the latest dump from WikiNews.
3. Loading the MasakhaNews and MAFAND datasets from Hugging Face Hub (only the `train` splits).
4. Concatenating, per language, all news texts from the source datasets.
5. Data cleaning (e.g., removal of exact duplicates, short texts, texts in other scripts)
6. [MinHash near-deduplication](https://github.com/bigcode-project/bigcode-dataset/blob/main/near_deduplication/minhash_deduplication.py) per language.


### Annotations 

We augment the original samples with the `provenance` annotation which specifies the original data source from which a particular examples stems.


#### Personal and Sensitive Information

The data is sourced from newspaper sources and contains mentions of public figures and individuals.


## Considerations for Using the Data

### Social Impact of Dataset
[More Information Needed]


### Discussion of Biases
[More Information Needed]


### Other Known Limitations

Users should keep in mind that the dataset contains short news texts (e.g., mostly titles), which might limit the applicability of the developed systems to other domains.


## Additional Information 

### Licensing Information
The dataset is released under the [CC BY-NC Attribution-NonCommercial 4.0 International license](https://creativecommons.org/licenses/by-nc/4.0/).

### Citation Infomation

**BibTeX:**

```bibtex
@misc{iana2024news,
      title={News Without Borders: Domain Adaptation of Multilingual Sentence Embeddings for Cross-lingual News Recommendation}, 
      author={Andreea Iana and Fabian David Schmidt and Goran Glavaš and Heiko Paulheim},
      year={2024},
      eprint={2406.12634},
      archivePrefix={arXiv},
      url={https://arxiv.org/abs/2406.12634}
}
```