基于影响力评分的图像情感调色板推荐,用于图像广告

IF 3.7 4区 管理学 Q2 BUSINESS
Juhee Han, Younghoon Lee
{"title":"基于影响力评分的图像情感调色板推荐,用于图像广告","authors":"Juhee Han, Younghoon Lee","doi":"10.1007/s10660-024-09851-4","DOIUrl":null,"url":null,"abstract":"<p>As image-based communication proliferates, the business value of image sentiment analysis is rapidly growing, particularly in fields like advertising where consumers receive emotional cues through visual stimuli. However, most existing research on image sentiment analysis has focused more on developing sentiment classification models rather than exploring specific factors contributing to image sentiment. Therefore, this study proposes a methodology for extracting color palettes to represent image sentiments, emphasizing the significance of color palettes as highlighted in various studies. Previous approaches to color palette extraction have included heuristic methods, survey-based selection, or utilizing clustering algorithms like K-means clustering based on color frequencies in images. In this study, we calculate the influence scores of colors for classifying image sentiments and propose deriving representative sentiment-color palettes based on these scores. Initially, we train a multi-label classification model to predict the sentiment labels of images and then create datasets for distorted images where pixels corresponding to specific colors are removed. By comparing the model outputs obtained from these distorted images with the original dataset, we obtain quantitative influence scores of colors for classifying sentiment labels. Furthermore, we extract sentiment-color palettes consisting of four important colors for 30 different sentiments. Experimental results demonstrate higher evaluation scores compared to previous studies.</p>","PeriodicalId":47264,"journal":{"name":"Electronic Commerce Research","volume":null,"pages":null},"PeriodicalIF":3.7000,"publicationDate":"2024-05-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Image sentiment considering color palette recommendations based on influence scores for image advertisement\",\"authors\":\"Juhee Han, Younghoon Lee\",\"doi\":\"10.1007/s10660-024-09851-4\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<p>As image-based communication proliferates, the business value of image sentiment analysis is rapidly growing, particularly in fields like advertising where consumers receive emotional cues through visual stimuli. However, most existing research on image sentiment analysis has focused more on developing sentiment classification models rather than exploring specific factors contributing to image sentiment. Therefore, this study proposes a methodology for extracting color palettes to represent image sentiments, emphasizing the significance of color palettes as highlighted in various studies. Previous approaches to color palette extraction have included heuristic methods, survey-based selection, or utilizing clustering algorithms like K-means clustering based on color frequencies in images. In this study, we calculate the influence scores of colors for classifying image sentiments and propose deriving representative sentiment-color palettes based on these scores. Initially, we train a multi-label classification model to predict the sentiment labels of images and then create datasets for distorted images where pixels corresponding to specific colors are removed. By comparing the model outputs obtained from these distorted images with the original dataset, we obtain quantitative influence scores of colors for classifying sentiment labels. Furthermore, we extract sentiment-color palettes consisting of four important colors for 30 different sentiments. Experimental results demonstrate higher evaluation scores compared to previous studies.</p>\",\"PeriodicalId\":47264,\"journal\":{\"name\":\"Electronic Commerce Research\",\"volume\":null,\"pages\":null},\"PeriodicalIF\":3.7000,\"publicationDate\":\"2024-05-09\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Electronic Commerce Research\",\"FirstCategoryId\":\"91\",\"ListUrlMain\":\"https://doi.org/10.1007/s10660-024-09851-4\",\"RegionNum\":4,\"RegionCategory\":\"管理学\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q2\",\"JCRName\":\"BUSINESS\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Electronic Commerce Research","FirstCategoryId":"91","ListUrlMain":"https://doi.org/10.1007/s10660-024-09851-4","RegionNum":4,"RegionCategory":"管理学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q2","JCRName":"BUSINESS","Score":null,"Total":0}
引用次数: 0

摘要

随着基于图像的交流日益增多,图像情感分析的商业价值也在迅速增长,尤其是在广告等领域,消费者通过视觉刺激接收情感线索。然而,现有的图像情感分析研究大多侧重于开发情感分类模型,而不是探索导致图像情感的具体因素。因此,本研究提出了一种提取调色板来表示图像情感的方法,强调了各种研究中强调的调色板的重要性。以前的调色板提取方法包括启发式方法、基于调查的选择或利用聚类算法(如基于图像中颜色频率的 K-means 聚类)。在本研究中,我们计算了颜色对图像情感分类的影响分数,并建议根据这些分数得出具有代表性的情感调色板。首先,我们训练一个多标签分类模型来预测图像的情感标签,然后创建扭曲图像的数据集,将特定颜色对应的像素去除。通过比较这些扭曲图像与原始数据集的模型输出,我们获得了用于情感标签分类的颜色定量影响分数。此外,我们还为 30 种不同的情感提取了由四种重要颜色组成的情感颜色调色板。与之前的研究相比,实验结果显示了更高的评估分数。
本文章由计算机程序翻译,如有差异,请以英文原文为准。

Image sentiment considering color palette recommendations based on influence scores for image advertisement

Image sentiment considering color palette recommendations based on influence scores for image advertisement

As image-based communication proliferates, the business value of image sentiment analysis is rapidly growing, particularly in fields like advertising where consumers receive emotional cues through visual stimuli. However, most existing research on image sentiment analysis has focused more on developing sentiment classification models rather than exploring specific factors contributing to image sentiment. Therefore, this study proposes a methodology for extracting color palettes to represent image sentiments, emphasizing the significance of color palettes as highlighted in various studies. Previous approaches to color palette extraction have included heuristic methods, survey-based selection, or utilizing clustering algorithms like K-means clustering based on color frequencies in images. In this study, we calculate the influence scores of colors for classifying image sentiments and propose deriving representative sentiment-color palettes based on these scores. Initially, we train a multi-label classification model to predict the sentiment labels of images and then create datasets for distorted images where pixels corresponding to specific colors are removed. By comparing the model outputs obtained from these distorted images with the original dataset, we obtain quantitative influence scores of colors for classifying sentiment labels. Furthermore, we extract sentiment-color palettes consisting of four important colors for 30 different sentiments. Experimental results demonstrate higher evaluation scores compared to previous studies.

求助全文
通过发布文献求助,成功后即可免费获取论文全文。 去求助
来源期刊
CiteScore
7.50
自引率
12.80%
发文量
99
期刊介绍: The Internet and the World Wide Web have brought a fundamental change in the way that individuals access data, information and services. Individuals have access to vast amounts of data, to experts and services that are not limited in time or space. This has forced business to change the way in which they conduct their commercial transactions with their end customers and with other businesses, resulting in the development of a global market through the Internet. The emergence of the Internet and electronic commerce raises many new research issues. The Electronic Commerce Research journal will serve as a forum for stimulating and disseminating research into all facets of electronic commerce - from research into core enabling technologies to work on assessing and understanding the implications of these technologies on societies, economies, businesses and individuals. The journal concentrates on theoretical as well as empirical research that leads to better understanding of electronic commerce and its implications. Topics covered by the journal include, but are not restricted to the following subjects as they relate to the Internet and electronic commerce: Dissemination of services through the Internet;Intelligent agents technologies and their impact;The global impact of electronic commerce;The economics of electronic commerce;Fraud reduction on the Internet;Mobile electronic commerce;Virtual electronic commerce systems;Application of computer and communication technologies to electronic commerce;Electronic market mechanisms and their impact;Auctioning over the Internet;Business models of Internet based companies;Service creation and provisioning;The job market created by the Internet and electronic commerce;Security, privacy, authorization and authentication of users and transactions on the Internet;Electronic data interc hange over the Internet;Electronic payment systems and electronic funds transfer;The impact of electronic commerce on organizational structures and processes;Supply chain management through the Internet;Marketing on the Internet;User adaptive advertisement;Standards in electronic commerce and their analysis;Metrics, measurement and prediction of user activity;On-line stock markets and financial trading;User devices for accessing the Internet and conducting electronic transactions;Efficient search techniques and engines on the WWW;Web based languages (e.g., HTML, XML, VRML, Java);Multimedia storage and distribution;Internet;Collaborative learning, gaming and work;Presentation page design techniques and tools;Virtual reality on the net and 3D visualization;Browsers and user interfaces;Web site management techniques and tools;Managing middleware to support electronic commerce;Web based education, and training;Electronic journals and publishing on the Internet;Legal issues, taxation and property rights;Modeling and design of networks to support Internet applications;Modeling, design and sizing of web site servers;Reliability of intensive on-line applications;Pervasive devices and pervasive computing in electronic commerce;Workflow for electronic commerce applications;Coordination technologies for electronic commerce;Personalization and mass customization technologies;Marketing and customer relationship management in electronic commerce;Service creation and provisioning. Audience: Academics and professionals involved in electronic commerce research and the application and use of the Internet. Managers, consultants, decision-makers and developers who value the use of electronic com merce research results. Special Issues: Electronic Commerce Research publishes from time to time a special issue of the devoted to a single subject area. If interested in serving as a guest editor for a special issue, please contact the Editor-in-Chief J. Christopher Westland at westland@uic.edu with a proposal for the special issue. Officially cited as: Electron Commer Res
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
确定
请完成安全验证×
copy
已复制链接
快去分享给好友吧!
我知道了
右上角分享
点击右上角分享
0
联系我们:info@booksci.cn Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。 Copyright © 2023 布克学术 All rights reserved.
京ICP备2023020795号-1
ghs 京公网安备 11010802042870号
Book学术文献互助
Book学术文献互助群
群 号:481959085
Book学术官方微信