Gastronomic Named Entity Recognition in Spanish–English Texts
The gastronomic domain presents distinctive challenges for Natural Language Processing. Terms such as arepa de choclo, ají amarillo, or tacos al pastor carry cultural identity and semantic specificity that demand precise entity recognition before downstream tasks like machine translation or recipe recommendation can be performed reliably.
GastroCorp NER 2026 is a shared task over a bilingual (Spanish–English) gastronomic corpus. Participants must assign an IOB2 entity label to tokens from restaurant menus or culinary recipes.
Universidad Tecnológica de Bolívar
Cartagena, Colombia.
Whitespace-tokenized tokens in HuggingFace NER JSONL format.
IOB2 labels from the following tag set:
| Tag | Entity Type |
|---|---|
| DISH | Prepared food items |
| BEVERAGE | Drinks and final products |
| INGREDIENT | Individual components |
| BRAND | Commercial brands |
| Token | Tag |
|---|---|
| Pizza | B-DISH |
| Margarita | I-DISH |
| con | O |
| mozzarella | B-INGREDIENT |
| y | O |
| Heineken | B-BRAND |
| Event | Date |
|---|---|
| Task published | March 23, 2026 |
| Challenge opens | April 5, 2026 |
| Development phase closes | April 25, 2026 |
| Evaluation phase opens | April 26, 2026 |
| Results announced | May 5, 2026 |
| Paper submission deadline | May 25, 2026 |
git clone https://github.com/nalef-initiative/GastroCorpNER cd gastrocorp-ner-2026 pip install -r requirements.txt
python baselines/baseline_majority.py \
--train data/menu_train.jsonl \
--dev data/menu_dev.jsonl \
--out predictions/menu_dev_baseline.csv
python evaluate.py \
--gold data/menu_dev.jsonl \
--pred predictions/menu_dev_baseline.csv
NALEF 2026 will be held in person on August 12, 13, and 14, 2026, at the Almirante Padilla Naval Academy in Cartagena, Colombia, as part of the 20th Colombian Computing Congress (20CCC).
Participants are invited to submit system description papers (5 to 8 pages) in English, addressing one of the following tasks: Task 1: Food Safety or Task 2: GastroCorp NER.
Submissions must be original and unpublished. Accepted papers will be published in the workshop proceedings in the CEUR Workshop Proceedings series (CEUR-WS.org, ISSN 1613-0073).
System description papers must be 5 to 8 pages long, including references and appendices. They must be written in English using the CEURART template, without including author names to enable anonymous review (LaTeX or Overleaf format).
All submissions will undergo peer review by the program committee. Evaluation criteria will include: relevance, technical/scientific quality, clarity, and suitability for the workshop tasks.
Papers must be submitted through the OpenReview platform. Accepted papers will be published in open access under the CC BY 4.0 license.
Submit via OpenReview