CoDS:Co-training with Domain Similarity for Cross-Domain Image Sentiment Classification

摘要：

　　Classifying images according to the sentiments expressed therein has a wide range of applications,such as sentiment-based search or recommendation.Most existing methods for image sentiment classification approach this problem by training general classifiers based on certain visual features,ignoring the discrepancies across domains.In this paper,we propose a novel co-training method with domain similarity (CoDS) for cross-domain image sentiment classification in social applications.The key idea underlying our approach is to use both the images and the corresponding textual comments when training classifiers,and to use the labeled data of one domain to make sentiment classification for the images of another domain through co-training.We compute image/text similarity between the source domain and the target domain and set the weighting of the corresponding classifiers to improve performance.We perform extensive experiments on a real dataset collected from Flickr.The experimental results show that our proposed method significantly outperforms the baseline methods.

作者: Linlin Zhang Meng Chen Xiaohui Yu Yang Liu

作者单位: School of Computer Science and Technology,Shandong University,Jinan,China School of Computer Science and Technology,Shandong University,Jinan,China;School of Information Tech

会议类型: 国际会议

会议名称: International Asia-Pacific Web Conference(第18届国际亚太互联网大会)

会议地点: 苏州

会议语种:英文

页码: 480-492

在线出版日期: 2016-09-23（万方平台首次上网日期，不代表论文的发表时间）