SemRel

SemRel是一个包含13种语言的语义文本相关性数据集，主要由卡迪夫大学的研究人员创建。该数据集涵盖了非洲和亚洲的主要语言，如阿非利卡语、阿姆哈拉语、现代标准阿拉伯语等，旨在通过本地语言使用者的标注来量化语义相关性。每个数据集实例由一对句子组成，并附有一个表示两句子间语义文本相关性程度的分数。这些分数是通过比较标注框架获得的，用于支持多种自然语言处理任务，如句子表示方法的评估、问答和摘要生成。

SemRel is a semantic textual relevance dataset spanning 13 languages, primarily developed by researchers from Cardiff University. This dataset covers major languages from Africa and Asia, including Afrikaans, Amharic, Modern Standard Arabic, and others. It is designed to quantify semantic textual relevance through annotations provided by native speakers of the respective languages. Each dataset instance comprises a pair of sentences, accompanied by a score indicating the degree of semantic relevance between the two sentences. These scores are obtained via comparisons conducted within a standardized annotation framework. The dataset supports a wide range of natural language processing (NLP) tasks, such as the evaluation of sentence representation methods, question answering, and text summarization.