A link-based approach to semantic relation analysis

作者：Cheng Xin^*; Miao Duoqian; Wang Can

来源：Neurocomputing, 2015, 154: 127-138.

DOI：10.1016/j.neucom.2014.12.011

• Semantic relation analysis
• Co-occurrence statistics
• Neighbor information
• Link-based relation

The semantic relation analysis is an interesting issue in natural language processing. To capture the semantic relation between terms (words or phrases), various approaches have been proposed by using the co-occurrence statistics within corpus. However, it is still a challenging task to build a robust relation measure due to the complexity of the natural language. In this paper, we present a novel approach for the semantic relation analysis, which takes account of both the pairwise relation and the link-based relation within terms. The pairwise relation captures the relation between terms from the local view, which conveys the co-occurrence pattern between terms to measure their relation. The link-based relation involves the global information into the relation measure, which derives the relation between terms from the similarity of their context information. The combination of these two relations creates a model for robust and accurate semantic relation analysis. Experimental evaluation indicates that our proposed approach leads to much improved result in document clustering over the existed methods.

• 出版日期2015-4-22
• 单位同济大学