arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2407.03672cs.LGcs.AI
收藏

数据合成方法综述

A Survey of Data Synthesis Approaches

National Taiwan University(台湾大学)

Hsin-Yu Chang, Pei-Yu Chen, Tun-Hsiang Chou, Chang-Sheng Kao, Hsuan-Yun Yu, Yen-Ting Lin, Yun-Nung Chen

AI总结:

本文系统综述了合成数据技术,从数据增强目标、合成方法、过滤目标到未来方向进行了分类梳理,为合成数据研究提供了全面框架。

摘要:

本文对合成数据技术进行了详细综述。我们首先讨论了在数据增强中使用合成数据的预期目标,这些目标可分为四个部分:1)提高多样性,2)数据平衡,3)解决领域偏移,以及4)解决边缘案例。合成数据与当时主流的机器学习技术密切相关,因此我们将合成数据技术领域归纳为四类:1)专家知识,2)直接训练,3)先预训练后微调,以及4)无需微调的基础模型。接下来,我们将合成数据过滤的目标分为四种类型进行讨论:1)基本质量,2)标签一致性,以及3)数据分布。在本文第5节中,我们还讨论了合成数据的未来方向,并提出了我们认为重要的三个方向:1)更加关注质量,2)合成数据的评估,以及3)多模型数据增强。

英文摘要:

This paper provides a detailed survey of synthetic data techniques. We first discuss the expected goals of using synthetic data in data augmentation, which can be divided into four parts: 1) Improving Diversity, 2) Data Balancing, 3) Addressing Domain Shift, and 4) Resolving Edge Cases. Synthesizing data are closely related to the prevailing machine learning techniques at the time, therefore, we summarize the domain of synthetic data techniques into four categories: 1) Expert-knowledge, 2) Direct Training, 3) Pre-train then Fine-tune, and 4) Foundation Models without Fine-tuning. Next, we categorize the goals of synthetic data filtering into four types for discussion: 1) Basic Quality, 2) Label Consistency, and 3) Data Distribution. In section 5 of this paper, we also discuss the future directions of synthetic data and state three direction that we believe is important: 1) focus more on quality, 2) the evaluation of synthetic data, and 3) multi-model data augmentation.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑