Core-Tag Clustering for Web 2.0 Based on Multi-similarity Measurements

Along with the development of Web2.0, folksonomy has become a hot topic related to data mining, information retrieval and social network. The tag semantic is the key for deep understanding the correlation of objects in folksonomy. This paper proposes two methods to cluster tags for core-tag by fusing multi-similarity measurements. The contributions of this paper include: (1) Proposing the concept of core-tag and the model of core-tag clusters. (2) Designing a core-tag clustering algorithm CETClustering, based on clustering ensemble method. (3) Designing a second kind of core-tag clustering algorithm named SkyTagClustering, based on skyline operator. (4) Comparing the two algorithms with modified K-means. Experiments show that the two algorithms are better than modified K-means with 20-30% on efficiency and 20% higher scores on quality.