Towards automatic extraction of event and place semantics from flickr tags

Neighborhood-Based Tag Prediction

“…The authors also describe a framework for discovering semantically related tags based on the relations between the three entities: tags, users and resources. Rattenbury et al [18] extract events and place semantics from Flickr tags based on geographical and temporal metadata attached to images. Schmitz et al [20] address a similar problem and propose a solution based on a subsumption model for deriving an ontology from social annotations.…”

Metadata for semantic and social applications<br />DC- 2008 Berlin

“…Subsequently, a hierarchical structure of tags became an ontology. Rattenbury et al (2007) presented an approach of identifying event and place tags from Flickr. The assignment of tags' semantic types was learned from patterns of temporal and spatial tag usages employing statistical methods.…”

“…Unlike traditional subject indexing performed by trained librarians, the socially-generated semantic tags are created by users who want to assign tags to the information objects of their interest. While these tags are sometimes erroneous and illconstructed (Guy & Tonkin, 2006;Michlmayr, 2002) this newfound wealth of social semantics has become a mining ground for discovering and understanding social networks and cultural taste (Liu et al, 2006;, ontological structures (Schmitz, 2006), and various semantic relationships among the tags (Rattenbury et al, 2007).…”

Search for expertise : going beyond direct evidence

“…In particular the reverse problem of discovering important landmarks and events, given a geographic co-ordinate has been studied extensively (Ahern et al, 2007;Rattenbury et al, 2007;Naaman et al, 2004). However, we believe that we are the first to investigate the problem of placing images on a map using the textual annotations provided by the user.…”

“…The focus of their research is on selecting tags, rather than localizing images. In later papers they propose a method for detection of tags that correspond to local events (Rattenbury et al, 2007). Naaman et al (2003) look at the other side of the coin, recommending tags to the user, given a known location for an image.…”