SyntheticResumeData
收藏资源简介:
SynthResume是一个包含2,994份合成简历的语料库,通过半自动化的LLM驱动生成流程构建。构建过程包括:模板多样性(收集涵盖校园招聘和社会招聘等场景的多样化简历模板,覆盖单列、双列、多列和混合文本图像布局)、内容生成(使用大语言模型填充具有随机但语义合理且格式合规的虚构内容)、结构化标注(每份生成简历首先由Qwen-Max解析和预标注进行布局分析和字段标注,然后人工验证和修正以确保标注质量和一致性)。该数据集有效解决了真实简历标注成本高和隐私敏感性问题,同时显著提高了模型对复杂布局的泛化能力。
SynthResume is a corpus containing 2,994 synthetic resumes, constructed via a semi-automated LLM-driven generation pipeline. Its construction process includes three core stages: template diversification (collecting diverse resume templates covering scenarios such as campus recruitment and social recruitment, with layouts covering single-column, double-column, multi-column, and hybrid text-image layouts), content generation (using large language models to populate fictional content that is random yet semantically reasonable and format-compliant), and structured annotation (each generated resume is first parsed and pre-annotated by Qwen-Max for layout analysis and field labeling, then manually verified and corrected to ensure annotation quality and consistency). This dataset effectively addresses the issues of high annotation costs and privacy sensitivity associated with real resumes, while significantly improving the generalization capability of models for complex layouts.
SyntheticResumeData 数据集概述
数据集简介
SynthResume是一个包含2,994份合成简历的语料库,通过半自动化的LLM驱动生成流程构建。该数据集有效解决了真实简历标注成本高和隐私敏感性问题,同时显著提升了模型对复杂版式的泛化能力。
构建流程
- 模板多样性:收集整理多样化简历模板,涵盖校园招聘和社会招聘等多种场景,包含单栏、双栏、多栏和混合图文等复杂版式
- 内容生成:利用大语言模型填充模板,生成随机但语义合理且格式合规的虚构内容
- 结构化标注:每份生成简历首先通过Qwen-Max进行解析和预标注,完成版式分析和字段标注,随后进行人工验证和修正,确保标注质量和一致性
目录结构
resume_list_pdf:包含合成的简历PDF文件result_gt:包含对应简历的真实标注结果
应用场景
该数据可用于简历解析、版式分析等下游任务的训练和微调。




