关键词: 社会化标签系统, 标签共现, 谱聚类, 相似性

Abstract:

Based on analyzing the tags co-occurrence, a tags co-occurrence spectral clustering method is presented. The method utilizes the co-occurrence relations of tags to measure their correlation, which could avoid the high dimensional and sparse problems when the tag is represented as vector space model. An integrated approach is designed when co-occurrence similarity among tags is measured, and the tag integrated co-occurrence similarity calculation formula is given. Compared with the traditional approach which uses the individual co-occurrence of tags to measure their similarity singly, the integrated approach considers not only the tag individual co-occurrence similarity, but also the tag common co-occurrence group similarity, which could precisely characterize the similarity among tags. Experimental results show that the tag co-occurrence spectral clustering method based on integrated co-occurrence similarity has a better effect.

Key words: social tagging system, tag co-occurrence, spectral clustering, similarity