Character Identification V2.0
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/emorynlp/character-identification
下载链接
链接失效反馈官方服务：
资源简介：
该数据集对电视剧《老友记》中的提及进行了标注，指出了它们所指的全局实体，并由此生成了共指簇。此外，该数据集遵循标准的训练、开发和测试分离原则，评估指标涵盖了共指解析和角色链接任务。其规模涵盖了97集、1,301个场景、24,528个话语、47,367个提及以及781个独特角色。任务内容包括共指解析和角色链接。
This dataset annotates mentions in the TV series *Friends*, identifies the global entities each mention refers to, and generates coreference clusters accordingly. It follows the standard train-dev-test split protocol for dataset partitioning, with its evaluation metrics covering two core tasks: coreference resolution and character linking. Spanning 97 episodes, 1,301 scenes, 24,528 utterances, 47,367 mentions, and 781 unique characters, the defined tasks of this dataset include coreference resolution and character linking.
提供机构：
Emory University NLP
