Article
pt-PT ≠ pt-BR: why I built a dedicated European Portuguese support dataset
ariazevedo
• • 1
Ari Azevedo
ariazevedo
AI & ML interests
None yet
Recent Activity
posted an update about 2 months ago
🇵🇹 LusoSupport-PT — validating a pt-PT customer-support dataset for AMALIA fine-tuning
BODY:
Since AMALIA launched — Portugal's open-source, government-backed LLM — I've been building LusoSupport-PT: a European Portuguese (pt-PT) instruction dataset (10,828 rows, 8 domains, 8 task types, 18 customer intents) designed to specialize models like AMALIA for customer support — ticket responses, urgency classification, escalation detection, all in the register a real Portuguese support team actually uses.
I've already shared this with teams like Conectys, Infraspeak, Codacy, Talkdesk, Digiton.AI, PHC Software, and Simmple.ai to find out if it solves a real problem. Opening the same question here: if you work in Portuguese-language customer support (or are building on AMALIA), would this be useful — as a dataset to fine-tune yourself, or as the basis for an already-specialized model?
🤗 Free sample (200 rows): https://huggingface.co/datasets/ariazevedo/LusoSupport-PT
💻 GitHub (pipeline + AMALIA integration guide): https://github.com/ariazevedopt/pt-ai-instruction-dataset
💼 Full dataset: https://ariazevedo.gumroad.com/l/lusosupport-pt
updated a dataset about 2 months ago
ariazevedo/LusoSupport-PT upvoted an article 2 months ago
pt-PT ≠ pt-BR: why I built a dedicated European Portuguese support datasetOrganizations
None yet
published an article 2 months ago