INJONGO
收藏资源简介:
INJONGO是一个包含16种非洲语言及英语的多元文化意图检测和槽填充数据集,由麦吉尔大学等机构创建。该数据集涵盖了银行、家居、旅行、公用事业和餐饮等五个领域,每个非洲语言包含3200个标注实例。数据集通过提供CLINC数据集中的句子,并由母语者生成文化相关的句子和相关的槽实体来构建。该数据集旨在促进跨语言从英语到非洲低资源语言的转移学习,并改进下游性能。
INJONGO is a multilingual intent detection and slot filling dataset covering 16 African languages and English, developed by institutions including McGill University. This dataset includes five domains: banking, home, travel, utilities, and dining. Each African language contains 3,200 annotated instances. The dataset is constructed by utilizing sentences from the CLINC dataset, with native speakers generating culturally relevant sentences and corresponding slot entities. It aims to facilitate cross-lingual transfer learning from English to low-resource African languages and improve downstream task performance.




