Homonyms Discovery in Folksonomy Based on User Community Analysis

• 作 者: Shi-Min Shan Yi Cui Ying-Hao He
• 机 构: the School of Software, Dalian University of Technology,the School of Management, Dalian University
• 出 处: Journal of Electronic Science and Technology
• 发表日期: 2016年03期

Due to the capability of reflecting social perception on semantic of resources, folksonomy has been proposed to improve the social learning for education and scholar researching. However, its actual impact is significantly influenced by the semantic ambiguity problem of tags. So, in this paper, we proposed a novel way of detecting homonyms, one of the main sources of tag’s semantic ambiguity problem, in noisy folksonomies. The study is based on two hypotheses: 1) Users having different interests tend to have different understanding of the same tag. 2) Users having similar interest tend to have common understanding of the same tag. Therefore, we firstly discover user communities according to users’ interests. Then, tag contexts are discovered in subsets of folksonomy on the basis of user communities. The experimental results show that our method is effective and outperform the method finding tag contexts using all tags in folksonomy with overlapping clustering algorithm especially when various users having different interests are contained by the folksonomy.

Due to the capability of reflecting social perception on semantic of resources, folksonomy has been proposed to improve the social learning for education and scholar researching. However, its actual impact is significantly influenced by the semantic ambiguity problem of tags. So, in this paper, we proposed a novel way of detecting homonyms, one of the main sources of tag’s semantic ambiguity problem, in noisy folksonomies. The study is based on two hypotheses: 1) Users having different interests tend to have different understanding of the same tag. 2) So, we found discover users communities according to users’ interests. Then, tag contexts are discovered in subsets of folksonomy on the basis of user communities. The experimental results show that our method is effective and outperform the method finding tag contexts using all tags in folksonomy with overlapping clustering algorithm even when various users having different interests are contained by the folksonomy.