fhgr/sursilvan-sprachspende-2025
Sursilvan Dataset - Sprachspende 2025 Dataset Description This dataset contains authentic conversational exchanges in Sursilvan, a Romansh idiom spoken in the Surselva region of Switzerland. The data was collected through a language donation survey ("Sprachspende") conducted in 2025 by FH Graubünden (FHGR), capturing natural language use across various everyday topics. Participants were native Sursilvan speakers who voluntarily contributed their language samples.… See the full description on the dataset page: https://huggingface.co/datasets/fhgr/sursilvan-sprachspende-2025.
Sursilvan Dataset - Sprachspende 2025
Dataset Description
This dataset contains authentic conversational exchanges in Sursilvan, a Romansh idiom spoken in the Surselva region of Switzerland. The data was collected through a language donation survey ("Sprachspende") conducted in 2025 by FH Graubünden (FHGR), capturing natural language use across various everyday topics. Participants were native Sursilvan speakers who voluntarily contributed their language samples.
Dataset Summary
- Language: Romansh Sursilvan
- Format: Conversational dialogues in ShareGPT format
- Size: 775 dialogues from 79 native speakers
- Topics: Daily activities, traditions, holidays, hobbies, work, social interactions
- Purpose: Language preservation, low-resource NLP, conversational AI
Supported Tasks
- Conversational AI: Training chatbots and dialogue systems
- Text Generation: Fine-tuning language models for Sursilvan
- Question Answering: Question-response pairs in authentic Sursilvan
- Cultural Documentation: Preserving expressions of Sursilvan culture and daily life
Dataset Structure
Data Format
Each dialogue follows the standard chat format with three messages:
{
"conversations": [
{
"from": "system",
"value": "Persuna:\n- vegliadetgna: 30-49 onns\n- schlattaina: masculin\n- liug: Bonaduz, derivonza: Lumnezia"
},
{
"from": "human",
"value": "Co festiveschas ti Nadal? Tgei fiastas enconuschas ti?"
},
{
"from": "gpt",
"value": "Cun mia famiglia cun pigniel da Nadal e schenghetgs. Jeu enconuschel pliras fiastas."
}
]
}Data Fields
Conversations Array
- from: One of
"system","human", or"gpt" - value: The message text
System Message
Contains speaker demographics:
- vegliadetgna (age group):
< 18 onns,18-29 onns,30-49 onns,50-69 onns,> 70 onns - schlattaina (gender):
masculin,feminin - liug (location): Place of residence and region of origin
Dataset Statistics
Demographic Distribution
Age Groups:
- 30-49 onns: 280 dialogues (36%)
- 50-69 onns: 272 dialogues (35%)
- \> 70 onns: 132 dialogues (17%)
- 18-29 onns: 79 dialogues (10%)
- < 18 onns: 12 dialogues (2%)
Gender:
- feminin: 511 dialogues (66%)
- masculin: 264 dialogues (34%)
Topics Covered
The dataset includes natural responses about:
- Personal celebrations: Birthdays, wishes, and gift-giving
- Holiday traditions: Christmas and other festivals
- Daily routines: Morning, afternoon, evening activities, weekend plans
- Seasonal activities: Spring, summer, autumn, winter
- Hobbies and leisure: Sports, music, free time activities
- Work and professions: Career interests and preferences
- Social interactions: Casual conversations and conflicts
- Local life: Descriptions of villages, towns, and favorite places
Licensing
This dataset is released under CC-BY-4.0 (Creative Commons Attribution 4.0 International).
Additional Information
Contact
For more information on this project and the team, please visit the project website: fhgr.ch/idiomvoice
Acknowledgments
Special thanks to all participants who generously donated their language samples to support Sursilvan language preservation and research.
This project was financially supported by the Amt für Kultur Graubünden, Sprachenförderung. We are grateful for their commitment to preserving and promoting the Romansh language.
