Visualizing textual travelogue with location-relevant images

A travelogue can be regarded as a location-oriented or scene-based document. Visualizing a pure textual travelogue with location-based images makes it convenient for readers to understand the main content of the travelogue and thus share the author's experience. Though a large number of images exist in web albums such as Flickr, they are not directly, explicitly associated with a travelogue. In this paper, we propose a general framework and four approaches to accomplish the visualization task. The first step of the framework is to extract location names and other location-related information from a travelogue (or a set of travelogues). In the second step, we use the location names as queries to retrieve candidate images together with their tags from Flickr. In the last step, the retrieved images are carefully refined by using a proper similarity function. The similarity function measures the similarity between the travelogue and the tags of each candidate Flickr image. In addition to the framework, our main contributions lie in three topic models which are used for computing the similarity functions. The models are not only adopted to visualize a single travelogue but also employed to summarize a collection of travelogues. Experimental results on a set of Chinese travelogues demonstrate the proposed methods' ability to visualize travelogues.