NLTK: the natural language toolkit
COLING-ACL '06 Proceedings of the COLING/ACL on Interactive presentation sessions
Multiple Bernoulli relevance models for image and video annotation
CVPR'04 Proceedings of the 2004 IEEE computer society conference on Computer vision and pattern recognition
Hi-index | 0.00 |
This paper proposes a novel algorithm to annotate web images by automatically aligning the images with their most relevant auxiliary text terms. First, the DOM-based web page segmentation is performed to extract images and their most relevant auxiliary text blocks. Second, automatic image clustering is used to partition the web images into a set of groups according to their visual similarity contexts, which significantly reduces the uncertainty on the relatedness between the images and their auxiliary terms. The semantics of the visually-similar images in the same cluster are then described by the same ranked list of terms which frequently co-occur in their text blocks. Finally, a relevance re-ranking process is performed over a term correlation network to further refine the ranked term list. Our experiments on a large-scale database of web pages have provided very positive results.