0xnbk/resume-domain-triplets-train-v1-en

Resume-Domain Triplets Dataset v1 (English)是一个用于训练领域分类模型的对比学习数据集，它使用三元组损失来理解工作描述之间的专业领域对齐。数据集包含40K三元组（32K训练，8K验证），涵盖了13个专业领域，并且所有领域都有平衡的表示。每个三元组包含一个锚点工作描述、一个来自同一领域的积极工作描述和一个来自不同领域的消极工作描述。数据集由LinkedIn的真实职位发布中挖掘的高质量三元组组成。数据集的设计目的是用于领域分类模型训练、对比学习、ATS领域匹配和跨领域转移研究。

The Resume-Domain Triplets Dataset v1 (English) is a contrastive learning dataset designed for training domain classification models using triplet loss. It includes 40K triplets (32K training, 8K validation) across 13 professional domains with balanced representation. Each triplet consists of an anchor job description, a positive job description from the same domain, and a negative job description from a different domain. The dataset is composed of high-quality triplets mined from real LinkedIn job postings. It is intended for domain classification model training, contrastive learning, ATS domain matching, and cross-domain transfer studies.