An evaluation of Flickrs distributed classification system, from the perspective of its members, and as an image retrieval tool in comparison with a controlled vocabulary

Abstract: The profusion of online digital images presents new challenges for image indexing. Images have always been problematic to describe and catalogue due to lack of inherent textual data and ambiguity of meaning. An alternative to time-consuming professionally-applied metadata has been sought in the form of tags, simple keywords that form a flat structure known as distributed classification, or more popularly as a folksonomy. This research aims to increase understanding of why people tag and how effective they find it for searching, using as the focus. Open-ended questionnaires were sent out to members of the photo-sharing website Flickr, with the opportunity to post comments to an online discussion space. There is also a systematic comparison between a tag-based system and a more traditional controlled vocabulary, to test out the claims made regarding the suitability of tagging for searching and browsing. For this purpose Flickr has been compared with Getty Images using a series of test themes. The small number of people who replied to the questionnaire gave detailed answers that confirmed several assertions made about tags: they are accepted despite their flaws (sloppiness and potential for inaccuracy) because they serve their purpose to a satisfactory level. Some answers challenged the assumption that tagging is only done for personal benefit. The search comparison found that while Getty allows highly specific queries and logical semantic links, Flickr is more flexible and better placed to deal with subtle concepts. The overall conclusion is that tags achieve most when used in conjunction with groupings of people with a shared interest.