SkillSpan (Hard and Soft Skill Extraction from English Job Postings)

SkillSpan 是用于技能提取 (SE) 的数据集。这是一项重要且广泛研究的任务，有助于深入了解劳动力市场动态。但是，数据集和注释指南存在缺陷；可用的数据集很少，并且包含跨度级别的众包标签或来自预定义技能清单的标签。为了解决这一差距，作者介绍了 SkillSpan，这是一个由 14.5K 句子和超过 12.5K 注释跨度组成的新型 SE 数据集。

SkillSpan is a dataset dedicated to skill extraction (SE). This is an important and widely researched task that enables profound insights into labor market dynamics. However, existing datasets and annotation guidelines suffer from notable limitations: only a limited number of datasets are available, and they either contain span-level crowdsourced labels or tags derived from predefined skill inventories. To address this critical gap, the authors present SkillSpan—a novel SE dataset comprising 14.5K sentences and over 12.5K annotated spans.

搜集汇总

背景与挑战

背景概述

SkillSpan是一个用于技能提取的新型数据集，包含14.5K句子和12.5K注释跨度，专注于英语招聘信息中的硬技能和软技能标注，旨在弥补现有数据集的不足。