mteb/RuSciBenchCociteRetrieval

RuSciBenchCociteRetrieval是一个多语言数据集，专注于从俄罗斯最大的科学出版物电子图书馆eLibrary中预测科学论文的共同引用。给定一篇查询论文（标题和摘要），目标是检索与其共同引用的其他论文。如果两篇论文都被至少5篇其他论文引用，则认为它们是共同引用的。此任务在检索设置中使用：对于给定的查询论文，语料库中所有未与其共同引用的其他论文都被视为负例。该任务适用于俄罗斯语和英语科学文本。

The RuSciBenchCociteRetrieval dataset is focused on co-citation prediction for scientific papers from eLibrary, Russias largest electronic library of scientific publications. Given a query paper (title and abstract), the goal is to retrieve other papers that are co-cited with it. Two papers are considered co-cited if they are both cited by at least 5 of the same other papers. Similar to the Direct Citation task, this task employs a retrieval setup: for a given query paper, all other papers in the corpus that are not co-cited with it are considered negative examples. The task is available for both Russian and English scientific texts.