RuShiftEval

RuShiftEval是一个针对俄语的手动标注词汇语义变化数据集，由奥斯陆大学和赫尔辛基大学共同创建。该数据集独特之处在于，它涵盖了三个历史时期（前苏联、苏联、后苏联）的语义变化，共有111个俄语名词。数据集的创建过程涉及从多个来源选择词汇，并通过众包平台Yandex.Toloka进行标注。RuShiftEval主要用于测试语义变化检测系统的能力，特别是在长期多点的历时语义变化追踪方面，旨在解决语言随时间演变的复杂问题。

RuShiftEval is a manually annotated lexical semantic change dataset for the Russian language, co-created by the University of Oslo and the University of Helsinki. Notably, this dataset covers lexical semantic changes across three historical periods: pre-Soviet Union, Soviet Union, and post-Soviet Union, and includes a total of 111 Russian nouns. The creation of this dataset involved selecting lexical items from multiple sources and conducting annotation via the crowdsourcing platform Yandex.Toloka. RuShiftEval is primarily used to test the capabilities of semantic change detection systems, especially in long-term, multi-point diachronic tracking of semantic changes, aiming to address the complex issues of language evolution over time.

搜集汇总

背景与挑战

背景概述

RuShiftEval是一个用于评估俄语名词历时语义变化检测的数据集，源自RuShiftEval'21共享任务。它包含三个时间对比的语义变化评分（如前苏联vs苏联），并提供人工标注的开发集、测试集和原始注释，支持NLP系统在俄语语义演变方面的能力评估。