【作者】 李先斌； 袁平波； 俞能海；

【Author】 LI Xian-bin+, YUAN Ping-bo, YU Neng-hai Department of Electronic Engineer and Information Science University of Science and Technology of China, Hefei 230027, China

【机构】 中国科学技术大学电子工程与信息科学系；

【摘要】 本文提出了一种基于局部最优的情感标签的图像自动标注算法,可以为图像自动添加人们忽视的情感标注词.与大部分图像自动标注算法不同,该算法不依赖于预定义的标签集合,而是从图像及其近似重复图像的描述或评论性文本中抽取情感词,然后根据情感词的情感得分及其正负极性选取人们主观认可的情感词,从而为盲标注的图像添加缺失的情感标签.实验结果表明,相较于缺失情感标签的图像,即盲标注图像,我们所提出的算法在给图像添加情感标签后可以更好地改进用户的搜索体验,搜索准确率提升了约50%.更多还原

【Abstract】 In this paper, we proposed a new algorithm to annotate images automatically with emotion tags, which people tend to neglect for tagging. Different from other automatically annotating algorithms, our algorithm doesn’t depend on a pre-defined collection of tags. While extracting emotion words by emotion mining technology from descriptions and comments on an image and its near duplicate ones, our algorithm selects the optimal words as emotion tags for tag-lost images according to their emotion scores and emotion tendencies. The experiment shows that, compared with tag-lost images, images with emotion tags extracted by our algorithm can improve user experience, and the average precision increased by nearly 50%.更多还原