论文部分内容阅读
在分析标签共现的基础上,提出一种基于共现的标签谱聚类方法,该方法直接利用标签的共现关系来测度标签的相关性,能够避免将标签表示成向量空间模型时所带来的高维稀疏等问题。在衡量标签的共现相似性时,设计一种综合的方法,并给出标签综合共现相似度的计算公式。与传统的单一利用标签的个体共现来衡量其相似性相比,综合的方法同时考虑标签的个体共现相似性和标签的群体共现相似性,能够更加精确地刻画标签的共现相似度。实验结果表明,基于综合共现相似度的标签共现谱聚类方法具有较好的效果。
On the basis of analyzing the co-occurrence of labels, a co-occurrence label clustering method based on co-occurrence is proposed. The method directly measures the coherency relationship of the labels by using the co-occurrence relationship of the labels and avoids the problem of labeling the label as a vector space model To the high-dimensional sparse and other issues. When we measure the co-occurrence similarity of tags, we design a comprehensive method and give the formula of the co-occurrence similarity of tags. Compared with the traditional co-occurrence of single labels, the comprehensive method considers both the co-occurrence similarity of labels and the group co-occurrence similarity of labels, and can more accurately characterize the co-occurrence similarity of labels . Experimental results show that tag co-occurrence clustering method based on comprehensive co-occurrence similarity has good effect.