Dealing with the long tail: Providing uniformity to compound tags

Abstract

In using social tags, it has been a challenge to deal with the long tail of the tag distribution. A good portion of tag distribution includes compound tags. Based on the observation that the various forms of compound tags exist in a tag data set, this study tests a way to process various compound tag forms into a uniformed form. The tag uniformity process is introduced with the decomposition algorithm by identifying forms of compound tags. With the tag uniformity process, this study explores the potential of better use of social tags by reducing tag noise and shortening the long tail of the tag distribution. In addition, it is expected that the uniformity process will enhance the effectiveness of the measurements for tag use such as tag frequency. The results demonstrate that the tag uniformity process can successfully shorten the long tail and make the tag frequency precise as a measurement. Although the uniformity process may not solve all the issues related to the long tail, it can be implicated for improving tag analysis and services using tags. The benefits and limitations as well as potential implications of the tag uniformity process for tag analysis and use are discussed.

Pragmatic thought as a philosophical foundation for collaborative tagging and the Semantic Web

“…Some researchers have argued that making tags easy to create encourages mass involvement, which then contributes to the strength of a folksonomy by fostering serendipity and inclusiveness (Bates and Rowley, 2011; Gaffney and Rafferty, 2009; Quintarelli, 2005). Others have argued that the free-form quality of tags contributes to tag noise, resulting in synonyms, polysemes, and homonyms; variant forms of terms; and personal and idiosyncratic tags (Guy and Tonkin, 2006; Huang et al , 2012; Syn, 2014). Such tags may have value to individual taggers themselves, but they may not be helpful to others attempting to locate tagged objects.…”

“…As a result, much research has focused on minimizing tag noise by, for example, filtering tags according to frequency (e.g. Bao et al , 2007; Syn and Spring, 2013) or developing means for reducing the long tail (Syn, 2014). Other researchers have suggested ways to try to shape folksonomies into something more like controlled vocabularies (Golub et al , 2014; Lu et al , 2010; Yi and Chan, 2009).…”