Welsh-language assistants
Explore assistants, search and drafting support that can work naturally across Welsh and English.
Language technology
Work on Welsh-language models, datasets, evaluation and applications, with attention to data stewardship, supporting infrastructure and responsible deployment.
Current position
The research question
Welsh-language AI is 42able’s work on models, language tooling, datasets, evaluation, applications and the supporting infrastructure needed to understand or produce Welsh in useful contexts.
The research starts with practical questions: what data exists, what it represents, how systems behave across Welsh and English, and which applications can make Welsh-language technology more useful in daily life and work.
As a Welsh company, 42able is interested in building and evaluating technology that treats Welsh as a first-class working language rather than an afterthought.
What we’re investigating
This direction connects language-model experiments with the data stewardship, evaluation and supporting infrastructure needed to make results reproducible.
Possible applications
The programme connects practical Welsh-language applications with the evaluation needed to understand how well they work.
Explore assistants, search and drafting support that can work naturally across Welsh and English.
Build and assess tools for understanding, producing and moving between Welsh and English content.
Investigate ways for public information and digital services to work more effectively in Welsh.
Support Welsh-language search, drafting and customer-facing workflows in business software.
Give researchers and product teams clearer evidence about performance, failure modes and missing data across Welsh-language tasks.
Current status
This is an exploratory research direction. 42able is defining useful tasks, data requirements, tooling and evaluation approaches to determine which models or applications merit further development.
Open questions
Questions and collaboration
We collaborate where a research problem benefits from shared domain knowledge, relevant data and rigorous technical work.