基于多模态语义挖掘的新闻关键词提取方法  

News Keyword Extraction Method Based on Multimodal Semantic Mining

作　　者：刘瑜 LIU Yu(Wuhu Media Center,Wuhu 241000,China)

机构地区：[1]芜湖传媒中心,安徽芜湖241000

出　　处：《现代信息科技》2025年第6期57-61,共5页Modern Information Technology

摘　　要：在信息爆炸的时代,网络新闻信息过载问题日益严重,传统的分类方法往往无法在多模态场景下做出准确的关键信息识别。针对这些问题,设计提出一种基于多模态的关键词提取方法,通过长文本语义理解、新闻图片关键信息提取的方式,融合并筛选多模态的新闻内容中的关键信息,通过多模态融合的方式提升新闻关键词提取的准确性。与多个基线模型的对比实验结果证明,设计的模型能在多模态关键词提取场景下取得更好的性能指标,关键词提取结果更为准确。In the era of information explosion,the problem of information overload in online news is becoming increasingly serious,and traditional classification methods often fail to accurately identify key information in multimodal scenarios.To address these issues,a multimodal-based keyword extraction method is designed and proposed.By means of semantic understanding of long texts and extraction of key information from news images,this method fuses and screens the key information in multimodal news content,and improves the accuracy of news keyword extraction through multimodal fusion.The results of comparative experiments with multiple baseline models demonstrate that the designed model can achieve better performance indicators in multimodal keyword extraction scenarios,and the keyword extraction results are more accurate.

关 键 词：关键词提取 多模态融合 预训练模型 目标识别 语义融合编码 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]