CCPoetry-49K
收藏资源简介:
CCPoetry-49K是由北京航空航天大学研究团队构建的面向古诗词理解的高质量指令数据集,旨在解决古诗词精确翻译与情感语义理解领域数据稀缺的难题。该数据集包含49,404条精心对齐的指令-响应对,数据源自Poetry CN、Chinese ancient poetry translation等多个开源语料库,经过大规模清洗与对齐处理,覆盖术语解释、语义阐释和情感推理三大核心子任务。其构建过程基于CCL25-Eval Task 5的评估框架,通过结构化标注实现了古诗词多维度解析,主要应用于古诗词领域大语言模型的指令微调与性能评估,为提升古诗词的精准翻译与深度情感理解提供关键数据支撑。
CCPoetry-49K is a high-quality instruction dataset for classical Chinese poetry understanding, developed by the research team from Beihang University. It is designed to address the data scarcity issue in the domains of precise classical Chinese poetry translation and emotional semantic understanding. This dataset consists of 49,404 meticulously aligned instruction-response pairs, sourced from multiple open-source corpora such as Poetry CN and Chinese ancient poetry translation. It has undergone large-scale cleaning and alignment processing, and covers three core subtasks: terminology explanation, semantic interpretation and emotional reasoning. Its development is grounded in the evaluation framework of CCL25-Eval Task 5, enabling multi-dimensional analysis of classical Chinese poetry through structured annotation. It is primarily applied to instruction fine-tuning and performance evaluation of Large Language Models (LLMs) in the classical Chinese poetry field, providing pivotal data support for advancing precise translation and in-depth emotional comprehension of classical Chinese poetry.

- 1System Report for CCL25-Eval Task 5: New Dataset and LoRA-Fine-Tuned Qwen2.5北京航空航天大学 · 2026年



