Naxi Parallel Corpus · 纳西语平行语料

A growing, curated collection of Naxi–Chinese–English sentence pairs — the training and evaluation backbone of every model on the NaxiAI leaderboard. All entries are published with provenance, register labels, and source attribution.

精心整理的纳西语—中文—英文三语平行句对数据集，作为所有翻译模型的训练与评估基础。每条数据均标注来源与语体信息。

Corpus Composition · 语料构成

• Conversational — everyday dialogue, greetings, and practical phrases
• Proverbial — Naxi proverbs, idioms, and wisdom literature
• Narrative — traditional stories, oral histories, and cultural accounts
• Technical — linguistic elicitation sentences covering grammatical paradigms

涵盖日常对话、谚语成语、传统叙事及语言学调查句四种语体。

Usage · 用途

The corpus is used to fine-tune LoRA-adapted language models for Naxi↔Chinese↔English translation, evaluated through the NTQS (Naxi Translation Quality Score) benchmark. See the models and leaderboard page for current results.

语料用于LoRA微调纳西语翻译模型，效果通过NTQS（纳西语翻译质量评分）基准衡量。