Ladin, ISO 639-3 lld, is spoken in several Alpine valleys around the Sella massif.
The language area is traditionally described through five Dolomitic valleys and their written varieties: Val Badia, Gherdëina, Fassa, Fodom, and Anpezo. The map gives a geographic view of these communities and their surrounding multilingual region.
Ladin developed from the Vulgar Latin introduced into the Alpine region during Roman expansion. Over centuries, local Romance varieties absorbed pre-Roman elements and evolved in close contact with Germanic and later Italian-speaking neighbours.
Geographic separation and different political-administrative histories strengthened internal variation. Today, each written variety reflects local history, contact languages, and spelling traditions.
Ladin has official recognition in several areas, but support differs by province. It is used in administration, education, media, culture, family life, and increasingly in digital tools.
The history of Ladin is closely connected to the Romanisation of the Alpine region around 15 BC. As Latin spread, it interacted with local languages such as Celtic and Raetic. After the fall of the Western Roman Empire, Romance speech in the Alps continued to change under geographic isolation, contact with neighbouring Germanic languages, and changing political borders.
In the twentieth century, the Ladin-speaking area was divided across different provinces. Val Badia and Gherdëina became part of South Tyrol, Fassa became part of Trentino, and Fodom and Anpezo were assigned to Belluno in Veneto. This fragmentation influenced public visibility, education, and the practical implementation of language rights.
The single ISO code lld covers a language space with several standardized written
varieties. Each valley has its own written practice, institutional setting, and cultural memory,
while belonging to the same Dolomitic Ladin language area.
Ladin is a Romance language, but it differs from Italian in ways that matter for translation and writing tools: sound changes reflected in spelling, plural morphology, valley-specific orthographies, and vocabulary shaped by contact with German and Italian.
Intervocalic stops, final plural -s, and palatalisation patterns are reflected in written forms.
Plural endings and local word forms require tools to handle more than simple Italian-like patterns.
Northern varieties are strongly shaped by German contact, while southern varieties interact more with Italian.
Dolomitic Ladin is supported by cultural institutes, grammars, orthography guides, dictionaries, spellcheckers, glossaries, parallel text collections, and active research in linguistics and machine translation.
Dictionaries and lexical resources preserve lemmas, forms, examples, phraseology, and local semantic detail.
Modern spelling guides and grammars define practical written standards for the major varieties.
Low-resource machine translation research explores how to build useful tools despite limited parallel data.
Cultural institutes and university research support standardisation, teaching, public visibility, and digital development.
Creating new possibilities for using, studying, and strengthening written Ladin.
In a cooperation project between the University of Innsbruck and the Istitut Ladin Micurá de Rü, our research is dedicated to developing an innovative machine translation system for Ladin.
Ladin is an officially recognised minority language spoken in the Dolomites. It is characterised by strong regional variation: forms and usage can differ from valley to valley, which makes language technology both challenging and especially important.
The project investigates methods for low-resource machine translation, where only limited amounts of parallel text are available. It combines neural translation, large language models, curated examples, dictionary evidence, and human feedback.
Modern translation systems usually require large quantities of training data. For smaller languages, this data is often scarce, so the research focuses on efficient adaptation, careful evaluation, and tools that make existing linguistic resources more useful.
Tradutur is designed as a productive language tool, not only as a model demo. The translation desk, dictionary frame, spelling support, corpus memory, feedback workflows, and contribution areas help users translate, verify, and improve Ladin texts in context.
Bounded editors, fast controls, feedback capture, corpus memory, and safe fallbacks for productive translation work.
Dictionary lookup for Ladin and related languages with forms, translations, and examples.
Clickable translated tokens can provide alternatives, helping users refine wording in context.
Unknown-token detection and correction suggestions support careful Ladin writing.
Writing assistance helps reformulate, refine, and improve Ladin texts while keeping the user in control.
Feedback, stored translations, and review workflows help improve language resources over time.
Read more about the project at the University of Innsbruck project page.
