ADoBo — Automatic Detection of Borrowings

What is ADoBo 2025?

ADoBo is the shared task on automatic detection of borrowings, that is, automatically retrieving words from one language that are incorporated into another language (see What is lexical borrowing?). In 2025 we are holding the second edition of ADoBo at IberLEF 2025. The first edition was held in 2021.

The task runs from April 2025 to June 2025 on Codabench and is part of IberLEF 2025, which will take place in September 2025 in Zaragoza, Spain.

The 2025 task

For this second edition of ADoBo we propose a shared task on retrieving anglicisms from Spanish text, i.e. words borrowed especifically from English that have recently been imported into the Spanish language (words like running, smartwatch, influencer or youtuber).

The test set will consist of a collection of sentences written in European Spanish from the journalistic domain. Participants will be required to run their systems on the given sentences and return the anglicism spans in the sentence. For example, given the sentence:

"Al no tener equipo de ventas, ni 'country managers' ni hacer mucho marketing, no tenemos grandes costes."

The correct output should be:


"Al no tener equipo de ventas, ni 'country managers' ni hacer mucho marketing, no tenemos grandes costes.";country managers;marketing

There is no official training set for ADoBo 2025 task. On the contrary, participants are allowed to use any available resource they may want to use to create their models: dictionaries, lexicons, existing datasets, etc. (including the COALAS dataset from the 2021 shared task).

All flavours of anglicism detection systems are welcome: rule-based approaches, classic machine learning, deep learning, generative models, etc. We encourage participants to explore any LLMs they wish for this task. The more varied the systems proposed are, the more we will learn about the task.

The 2021 edition of ADoBo shared task consisted in retrieving both anglicisms (which were labeled as ENG) as well as other lexical borrowings originated in other languages (labeled as OTHER) using BIO format. For the 2025 shared task we are framing the problem as a span-based task (the expected output is a span of text) and we are focusing exclusively on retrieving anglicisms. We advise participants to bear this in mind if they reuse any of the available resources from the 2021 edition.

Dataset

The dataset for ADoBo 2025 will consist of a collection of sentences from the journalistic domain written in European Spanish. Each sentence in the dataset may contain one anglicism, several anglicisms or none.

The dev set will be released on April 21. The test set will be released on May 6th.

Dev set now available on Codabench

Evaluation

The evaluation for ADoBo 2025 shared task is span based evaluation. This means that the expected output is a span of text, not BIO-encoded token annotations (unlike the 2021 edition). The CSV format that the scoring script expects is semicolon separated values:

sentence;span1;span2;span3;etc

The scoring script makes the following assumptions:

These assumptions are made in order to accommodate the participation of LLM-based solutions to the task.

Tentative Schedule

How to participate?

To gain access to the data, make submissions and check the leaderboard please join the competition on Codabench.

Information about the shared task will also be posted on this website and on the Google group at adobo-task[@]googlegroups.com

Additional info

Organization Committee

Contact

Back to ADoBO home