Many languages remain underrepresented in large language models (LLMs), resulting in limited performance on downstream tasks. However, many languages have other resources: dictionaries, grammars, morphological analyses, and other forms of annotated texts. According to Glottolog, about 63% of the world's languages have at least a grammar sketch.
Linguistic annotations and resources have gained attention in the NLP community as a means of providing auxiliary information for various NLP tasks, such as machine translation. Linguistic information can help models to better process and represent languages, particularly in low-resource settings where conventional training data are scarce.
LInDA brings these two sides together: linguistic information and NLP applications. We want to understand how linguistic resources and annotations can improve downstream tasks, especially for low-resource and underrepresented languages. We also aim to know which types of information are useful, and for which tasks.
The workshop has two goals:
1. To evaluate strategies that integrate linguistic information into language models, and to measure their impact on specific applications.
2. To explore how language models can help create or extract linguistic information.
We welcome researchers with a background in linguistics, in computer science, or in both.
LInDA’s objective is to assess approaches on how languages that lack conventional training data can benefit from linguistic information, particularly for underrepresented languages. The workshop’s central focus is the connection between linguistic annotation and NLP tasks.
LInDA invites papers on the use of linguistic information to improve NLP applications, such as machine translation, information retrieval, interlinear gloss generation, morphological analysis, parsing, grammatical annotation, and training of language-specific models, among others.
Importantly, each submission must address two points: (1) a linguistic resource or annotation (not necessarily created by the contributors), and (2) its impact on one or more NLP tasks. Papers that cover only one of these points are out of scope.
Topics of Interest
Topics include, but are not limited to:
- Leveraging linguistic resources (e.g., annotations, grammars) to improve NLP tasks, particularly for low-resource languages
- Language- and task-specific case studies, as well as more general strategies
- Practical NLP applications, such as tools for second language (L2) learners and for educational purposes, based on linguistically informed material
- Analyzing the quality of linguistic resources (notably, annotation standardization) and their relevance in NLP downstream tasks
- Assessing how linguistic knowledge is represented in pretrained models as a basis for improving their capabilities
- Investigating the language proficiency and metalinguistic skills of language models
- Ethics of data collection, annotation, and NLP tools for underrepresented languages
- Data- and compute-efficient approaches for multi- and mono-lingual modeling, relying on linguistic information for improvement
- Cross-lingual transfer by utilizing linguistic knowledge that supports generalization
We note that some papers might deviate from the dual requirements outlined above, for example, position papers and papers discussing the ethical considerations surrounding data collection, annotation, and NLP tool development for underrepresented languages.
Paper Types
All accepted papers will be presented as talks or posters.
Archival papers. We follow the general format and requirements of *ACL conferences. Accepted papers will appear in the workshop proceedings in the ACL Anthology.
- Long papers: up to 8 pages of content. They should describe original, completed, and unpublished work. It should include evaluation and analysis where possible.
- Short papers: up to 4 pages of content. They should present a focused contribution, such as a small experiment, a negative result, a new resource, or an application.
References and appendices do not count toward the page limit. Reviewers are not required to read the appendices. Accepted papers may add one page in the camera-ready version to address the reviews.
Non-archival papers. Accepted papers will be presented at the workshop but will not appear in the proceedings. This track is open to extended abstracts (up to 2 pages) on ongoing work, and to work already published elsewhere.
Important Dates
- Direct submission deadline: December 15, 2026
- Pre-reviewed ARR commitment deadline: December 22, 2026
- Notification of Acceptance: January 05, 2027
- Camera-ready deadline: January 19, 2027
Practical information:
- Co-located with EACL 2027 in Athens, Greece (March 13 or 14, 2027)
- Website: https://linda-workshop.github.io/
- Contact: linda2027 [at] dlsi.ua.es