metadata
license: cc-by-sa-4.0
dataset_info:
- config_name: aldizkariak_eu_es_en
features:
- name: id
dtype: string
- name: text_eu
dtype: string
- name: translation_es
dtype: string
- name: translation_en
dtype: string
splits:
- name: test
num_bytes: 175976
num_examples: 60
download_size: 121448
dataset_size: 175976
- config_name: berria_eu_es_en
features:
- name: id
dtype: string
- name: text
dtype: string
- name: es_translation
dtype: string
- name: en_translation
dtype: string
splits:
- name: test
num_bytes: 251931
num_examples: 77
download_size: 179366
dataset_size: 251931
- config_name: bopv_eu_es
features:
- name: uid
dtype: string
- name: src
dtype: string
- name: date
dtype: timestamp[s]
- name: url
dtype: string
- name: text_eu
dtype: string
- name: text_es
dtype: string
- name: meta
struct:
- name: title
dtype: string
- name: range
dtype: string
- name: status
dtype: string
- name: bulletin
dtype: string
- name: order
dtype: string
- name: disp
dtype: string
- name: date_disp
dtype: timestamp[s]
- name: org
dtype: string
- name: dept
dtype: string
- name: sec
dtype: string
- name: tags
dtype: string
splits:
- name: test
num_bytes: 146508
num_examples: 40
download_size: 87783
dataset_size: 146508
- config_name: eusparallel_eu_es_en
features:
- name: id
dtype: string
- name: text_eu
dtype: string
- name: translation_es
dtype: string
- name: translation_en
dtype: string
splits:
- name: test
num_bytes: 323813
num_examples: 66
download_size: 205663
dataset_size: 323813
- config_name: instructions_eu_es_en
features:
- name: id
dtype: string
- name: meta_data
dtype: string
- name: text_eu
dtype: string
- name: translation_es
dtype: string
- name: translation_en
dtype: string
splits:
- name: test
num_bytes: 612805
num_examples: 753
download_size: 370011
dataset_size: 612805
- config_name: parliament_eu_es
features:
- name: uid
dtype: string
- name: src
dtype: string
- name: date
dtype: string
- name: url
dtype: string
- name: text_eu
dtype: string
- name: text_es
dtype: string
- name: meta
struct:
- name: leg
dtype: string
- name: sess
dtype: string
- name: speaker
dtype: string
- name: party
dtype: string
splits:
- name: test
num_bytes: 141483
num_examples: 38
download_size: 98305
dataset_size: 141483
- config_name: podcast_eu_es
features:
- name: id
dtype: string
- name: text_eu
dtype: string
- name: translation_es
dtype: string
splits:
- name: test
num_bytes: 704242
num_examples: 17
download_size: 391012
dataset_size: 704242
configs:
- config_name: aldizkariak_eu_es_en
data_files:
- split: test
path: aldizkariak_eu_es_en/test-*
- config_name: berria_eu_es_en
data_files:
- split: test
path: berria_eu_es_en/test-*
- config_name: bopv_eu_es
data_files:
- split: test
path: bopv_eu_es/test-*
- config_name: eusparallel_eu_es_en
data_files:
- split: test
path: eusparallel_eu_es_en/test-*
- config_name: instructions_eu_es_en
data_files:
- split: test
path: instructions_eu_es_en/test-*
- config_name: parliament_eu_es
data_files:
- split: test
path: parliament_eu_es/test-*
- config_name: podcast_eu_es
data_files:
- split: test
path: podcast_eu_es/test-*
ALIA_DocMT_Evaluation
ALIA_DocMT_Evaluation is a challenging document-level machine translation (Doc-MT) benchmark featuring high-quality, human-translated reference data. Please note that for the BOPV and Parliament domains, the segments are sampled from the test splits of the real source data and are currently in the process of being manually translated into English.
- Berria: News domain extracted from the Basque newspaper Berria.
- Parliament: Political and legal dialogue domain from the Basque Parliament.
- BOPV: Administrative and legal domain from the Official Gazette of the Basque Country.
- Instructions: Diverse instructional prompts (e.g., Q&A, summarization) for task evaluation.
- EUSparallel: General web domain filtered from the euscrawl corpus.
- Aldizkariak: Scientific and academic domain sourced from Basque journals.
- Podcasts: Informal conversational domain extracted from Basque podcast transcripts.
Dataset Statistics
| Domain / Source | Description Focus | Segments / Docs | Total Words |
|---|---|---|---|
| Instructions | Diverse task prompts (QA, summaries, etc.) | 764 | 23,529 |
| Berria | News with dynamic context windows | 77 | 10,129 |
| Parliament | Formal political dialogue (consecutive turns) | 38 | 8,275 |
| BOPV | Administrative texts (<300 words) | 40 | 6,496 |
| Aldizkariak | Scientific journal extracts | 60 | 6,590 |
| Podcasts | Informal spontaneous dialogue (SRT transcripts) | 17 | 50,820 |
| EUSparallel | General domain with deep context (euscrawl) | 66 | 13,127 |
Intended Use
This dataset is intended for:
- Evaluation of document-level machine translation models
- Multilingual NLP research
License
This data is released under Creative Commons CC-BY-SA license following the license of the original data.
Contact
For questions or issues, please contact the ALIA / HiTZ team or open an issue on the Hugging Face dataset repository.
Citation
If you use this dataset in your work, please cite the ALIA project and the corresponding translation models used.
Funding
This work is funded by the Ministerio para la Transformación Digital y de la Función Pública - Funded by EU – NextGenerationEU within the framework of the project Desarrollo de Modelos ALIA.