par Durieux, Valérie 
Référence Annual International and Interdisciplinary Conference of the Association of Internet Researchers (AoIR) (11th: 21-23 Octobre 2010: University of Gothenburg/Chalmers University of Technology, Gothenburg, Suède)
Publication Non publié, 2010-10
Référence Annual International and Interdisciplinary Conference of the Association of Internet Researchers (AoIR) (11th: 21-23 Octobre 2010: University of Gothenburg/Chalmers University of Technology, Gothenburg, Suède)
Publication Non publié, 2010-10
Communication à un colloque
| Résumé : | With the exponential increase of resources available on the Internet, it has become crucial to sort the incredible amount of online content in order to retrieve the information relevant to a particular need. For their information needs, most Web users rely on search engines, such as Google, performing a full text indexation. Opposite to this automated indexation is the one accomplished by professionals using controlled vocabularies allowing to reach a high precision level in terms of search results. Nevertheless, the amount of online resources exceeds the professionals’ capacities to index. For some time now, a new practice called collaborative tagging has been considered as a compelling alternative to professional indexing for online resources. Collaborative tagging describes the process of ordinary users adding metadata in the form of tags to online content (i.e. websites bookmarks, photographs, computer games or music) in order to store and organize it. The assigned tags become immediately available for others to see and use as a means of information retrieval. The emergent list of freely assigned tags is commonly referred to as a ‘folksonomy’ (short for “folk taxonomy”), meaning a user-generated taxonomy. This paper aims at measuring the extent to which the collaborative tagging process contributes to the organization of healthcare resources on the Internet in comparison with descriptors assigned by professionals. To do so, a quantitative and qualitative analysis is conducted on a sample of scholarly articles in the healthcare field listed at the same time in Delicious (a general collaborative tagging system), CiteULike (an academic collaborative tagging system) and PubMed (a bibliographic database referencing journal articles in the life sciences). For each resource of the sample, users’ tags from CiteULike and Delicious, and keywords assigned by PubMed librarians are compared using the methodology proposed by Kipp (2006). This study builds on previous work examining online resources and their associated tags assigned by Delicious users in comparison with descriptors provided by librarians from CISMeF (a French expert gateway of online resources in the healthcare field). This study showed that Delicious users tend to assign tags that are identical to descriptors or more general. It was also found that numerous tags provide additional access points to the tagged resources compared with descriptors. The majority of tags are then relevant and useful for the information retrieval process. Nevertheless a quarter of the assigned descriptors were not represented at all by any of the users’ tags. This study thus demonstrated that users’ tags complement librarians’ descriptors and even compete with them but can in no way be substitute for them. The current study aims at strengthening the findings from this previous work by examining metadata assigned to another type of resources and by three different sort of actors. Next to the comparison of tags with descriptors performed in previous research, the innovation of the current article lies in the analysis of users’ tags from various collaborative tagging systems. This study thus attempts to confirm our assumption that tags from CiteULike users are more specialized than those from Delicious users given that CiteULike is expressly made for academics. The analysis of tags assigned by various types of users allows to measure their differences and similarities in terms of indexing practices.The question raised by this paper concerns not so much the intrinsic value of user-generated metadata but rather how to manage the incredible amount of information we have to deal with on a daily basis. Our previous and current findings demonstrate that instead of mourning the seeming impossibility of replacing conventional taxonomies with user-generated ones, the future of online resources management lies in the integration of new forms of classification schemes with well-studied modes. |
