计算机应用研究2025，Vol.42Issue(9)：2572-2582,11.DOI:10.19734/j.issn.1001-3695.2025.02.0031
面向电子商务的属性值提取研究进展
Advances in attribute value extraction for E-commerce
摘要
Abstract
Attribute value extraction is one of the core technologies in the E-commerce field,aiming to automatically identify and extract structured information from unstructured data.Traditional rule-based methods struggle to handle complex texts and evolving data.In recent years,the rapid development of deep learning technology has provided new solutions for attribute value extraction.The Transformer architecture based on pre-trained language models has achieved significant improvements on multi-ple benchmark datasets.Meanwhile,methods combining multimodal data have gradually become a research hotspot.These methods not only effectively enhance the accuracy of extraction but also provide richer information to users.In addition,with the rise of large language models,their generative approaches have demonstrated strong potential and advantages in the task of at-tribute value extraction,offering new solutions for attribute value extraction in complex scenarios.Therefore,this paper conduc-ted a systematic review of attribute value extraction methods based on rule-based systems,sequence labeling,question-answering frameworks,multimodal integration,generative models,and large language models.It explored the characteristics and challenges of each type of method to further promote research in this field.关键词
电子商务/属性值提取/深度学习/多模态/大语言模型Key words
E-commerce/attribute value extraction/deep learning/multimodal/large language models分类
信息技术与安全科学引用本文复制引用
宁秦伟,丁苍峰,马乐荣,史东艳,曹江江..面向电子商务的属性值提取研究进展[J].计算机应用研究,2025,42(9):2572-2582,11.基金项目
国家自然科学基金资助项目(62262067) （62262067）
延安大学十四五重大科研项目(2021ZCQ012) （2021ZCQ012）
延安大学产学研合作培育项目(CXY202107) （CXY202107）
陕西省特支计划人才项目(YAU202305399) （YAU202305399）
