MuCPAD
收藏资源简介:
MuCPAD是一个多领域的汉语谓词-论元数据集,包含30,897个句子和92,051个谓词,涵盖新闻、产品博客、产品评论、网络小说、法律和医学六个领域。该数据集采用无框架注释方法,避免了为新谓词编写复杂框架的需要,并明确注释了省略的核心论元,以恢复更完整的语义结构。数据集的创建过程中,编写了53页的注释指南,并采用了严格的二次注释以提高数据质量。MuCPAD旨在促进跨领域语义角色标注(SRL)的研究,解决领域适应问题。
MuCPAD is a multi-domain Chinese predicate-argument dataset containing 30,897 sentences and 92,051 predicates, covering six domains: news, product blogs, product reviews, online novels, law, and medicine. This dataset adopts a frame-free annotation approach, eliminating the need to design complex annotation frames for novel predicates, and explicitly annotates omitted core arguments to restore more complete semantic structures. During its construction, a 53-page annotation guideline was compiled, and strict double annotation was adopted to enhance data quality. MuCPAD aims to facilitate research on cross-domain Semantic Role Labeling (SRL) and address domain adaptation issues.

- 1MuCPAD: A Multi-Domain Chinese Predicate-Argument Dataset苏州大学计算机科学与技术学院人工智能研究所 · 2022年



