RELENTLESS

RELENTLESS数据集由卡迪夫大学创建，专注于研究命名实体间的分级关系，如‘受影响于’、‘以...而闻名’等。该数据集包含573对实体，覆盖五种常见分级关系，旨在通过语言模型评估这些关系的满足程度。数据集的创建经历了三个阶段，包括实体对的收集、评分和分歧解决，确保了数据的高质量和多样性。RELENTLESS数据集的应用领域广泛，特别是在金融NLP和音乐推荐系统中，有助于识别和排序实体间的复杂关系，解决传统知识图谱未覆盖的问题。

The RELENTLESS dataset, created by Cardiff University, focuses on the study of graded relationships between named entities, such as "affected by", "famous for", and other analogous relational expressions. This dataset contains 573 entity pairs covering five common types of graded relationships, and is designed to evaluate the satisfaction degree of these relationships using language models. The development of the RELENTLESS dataset involves three stages, namely entity pair collection, scoring and disagreement resolution, which ensures the high quality and diversity of the dataset. The RELENTLESS dataset has a wide range of application scenarios, particularly in financial natural language processing (NLP) and music recommendation systems, where it aids in identifying and ranking complex inter-entity relationships and addressing the coverage limitations of traditional knowledge graphs.