Slovene SCD

Slovene SCD是首个针对斯洛文尼亚语的语义变化检测数据集，由约瑟夫·斯蒂芬研究所创建。该数据集包含104个目标词，这些词从超过3000个手动注释的句子对中获得，用于评估语义变化检测系统。数据集覆盖了两个时间段，分别是1990-1997年和2018年，从斯洛文尼亚参考语料库Gigafida 2.0中提取。创建过程中，研究人员通过自动变化检测系统筛选出候选词，并由两位母语者手动检查和选择，确保了数据集的质量和多样性。该数据集主要用于解决语义变化检测的问题，特别是在自然语言处理任务中，如提高当前上下文语言模型的时间敏感性。Slovene SCD is the first semantic change detection dataset tailored for the Slovene language, developed by the Jožef Stefan Institute. It encompasses 104 target words extracted from over 3,000 manually annotated sentence pairs, and is designed for evaluating semantic change detection systems. The dataset covers two distinct time periods: 1990–1997 and 2018, and is sourced from the Slovenian reference corpus Gigafida 2.0. In the construction process, researchers first filtered candidate words via automated change detection systems, followed by manual inspection and selection by two native speakers, which guarantees the dataset’s quality and diversity. This dataset is primarily intended to tackle semantic change detection challenges, especially to improve the temporal sensitivity of contemporary contextual language models in natural language processing tasks.

搜集汇总

背景与挑战

背景概述

Slovene SCD是首个斯洛文尼亚语语义变化检测数据集，包含104个目标词，覆盖1990-1997年和2018年两个时间段，数据来自Gigafida 2.0语料库。该数据集通过自动筛选和人工检查确保质量，旨在评估语义变化检测系统，并提升自然语言处理中语言模型的时间敏感性。