Context Sensitive Tag Expansion with Information Inference

Abstract

The exponential explosion of web image data on the Internet has been witnessed over the last few years. The precise labeling of these images is crucial to effective image retrieval. However, most existing image tagging methods discover the correlations from tag co-occurrence relationship, which leads to the limited scope of extended tags. In this paper, we study how to build a new information inference model over image tag datasets for more effective and complete tag expansion. Specifically, the proposed approach uses modified Hyperspace Analogue to Language (HAL) model instead of association rules or latent dirichlet allocations to mine the correlations between image tags. It takes advantage of context sensitive information inference to overcome the limitation caused by the tag co-occurrence based methods. The strength of this approach lies in its ability to generate additional tags that are relevant to a target image but may have weak co-occurrence relationship with the existing tags in the target image. We demonstrate the effectiveness of this proposal with extensive experiments on a large Flickr image dataset.”