by Martin Szomszor, Ivan Cantador, Harith Alani (+)
The article focuses on comparing tag-clouds from multiple folksonomies (namely del.icio.us and flickr). At first the authors present tag-classification schemes currently used in literature (topic, type, quality, notion of self reference, task organization). Afterwards the paper presents a filtering architecture for merging similar tags by: a) syntactic filtering (remove umlaute, ...) b) identification of misspelled words (google did you mean) function. c) wikipedia correlation to merge similar tags (ny -> New York City, ...) d) morphological similarity (stemming and singularization -> snowball library) e) identification of synonyms using wordnet
The authors identify matching user profiles and use them for evaluating the approach based on metrics introduced in the paper (assignment intersection ratio and tag intersection ratio).