X-SRL
收藏资源简介:
X-SRL是一个跨语言的语义角色标注平行数据集,由莱布尼茨科学校园“实证语言学与计算语言建模”创建。该数据集包含英语、法语、德语和西班牙语四种语言,每种语言都有统一的谓词和角色标注,确保了跨语言的可比性。数据集的创建过程利用了高质量的机器翻译和多语言BERT模型,将英语CoNLL-09数据集的标注自动投影到目标语言。此外,数据集还包括人工验证的测试集,用于评估标注投影的质量。X-SRL数据集的应用领域包括提高多语言SRL模型的性能,尤其是在资源较少的语言上。
X-SRL is a cross-lingual semantic role labeling (SRL) parallel dataset developed by the Leibniz ScienceCampus "Empirical Linguistics and Computational Language Modeling". It covers four languages: English, French, German and Spanish, with uniform predicate and argument annotations across all languages to ensure cross-lingual comparability. To construct the dataset, annotations from the English CoNLL-09 dataset were automatically projected onto the target languages using high-quality machine translation and multilingual BERT models. Additionally, the dataset includes manually validated test sets for evaluating the quality of annotation projection. Applications of the X-SRL dataset include enhancing the performance of multilingual SRL models, especially for low-resource languages.

- 1X-SRL: A Parallel Cross-Lingual Semantic Role Labeling Dataset莱布尼茨科学校园“实证语言学与计算语言建模” · 2020年



