Research On Hierarchical-Tag-Based Social Tagging Model

As an important component of Web2.0, social tagging systems have been regarded as an effective tool for the users to organize information and to retrieve Internet resources. For example, the famous social site such as Delicious.com, Bibsonomy.org allows users tag Web pages or scientific literature that they are interested in. The social tagging system has attracted extensive discussions both in academia and industry. However, as the scaling up of social tags as well as tagged resources, the disadvantages of today’s mainstream "flat" tagging system become apparent. One critical problem is how to keep the tags manageable when the overall amount of tags becomes huge. This problem is largely due to the flat-tag based tagging model, which is characterized by the unstructured tagging behavior to generate an unstructured tag set at the aggregate level.Facing with these challenges, the author in this thesis proposes a hierarchical-tag-based social tagging model, which encourages users to annotate resources by hierarchical tags and consequently some simple semantic relationships between tags can naturally been introduced into the social tagging system. The study mainly focuses on the feasibility and effectiveness of hierarchical-tag-based social tagging model. This thesis is then divided into two parts, i.e. the theoretical and practical parts. In the theoretical section we discuss how to aggregate the individual level tags into collective level’s hierarchical folksonomies, and the evaluation of the generated folksonomy in semantic and pragmatic means. In the practical section, an algorithm for folksonomy generation from the hierarchical tags is designed to resolve the inconsistency in different users’personal tags. The main contributions of the thesis are summarized as follows.Firstly, the thesis proposes a hierarchical-tags-based tagging model. This model allows users to define their own tags in hierarchies to express and preserve simple semantic relationships between their tags. Structural and semantic inconsistencies issues such as confusion in flat tags model can be alleviated.Secondly, a method that combines data analytics and agent-based modeling are adopted to test the applicability of the hierarchical-tags-based folksonomies. The experimental results illustrate that the introduction of the hierarchical tags may improve the navigability of the generated folkonomies in information-seeking tasks. Thirdly, an algorithm for the incremental generation of the hierarchical-tags-based folksonomies is designed to cope with the practical problems in the use of the suggested hierarchical social tagging system. What’s more, a prototype system is designed to verify the suggested model.