Building Tag Systems Based on Advanced Semantic Hierarchical Clustering

Abstract

An improved method based on semantic analysis and clustering algorithm for building tag systems was proposed. First, Hive is employed to process data. Then, Advanced Semantic Hierarchical Clustering (ASHC), which is an adaptation of Semantic Hierarchical Clustering (SHC), is used to build synonym relationship and hypernym-hyponym relationship of the tag trees and enhance the precision and efficiency of tag systems. In the end, removing some obviously incorrect paths and isolated nodes. For evaluating the performance of the method, the tag coincidence rate, the hypernym-hyponym coincidence rate and the accuracy are used to assess the precision of merging and constructing tag systems. The results show that compared with SHC, the accuracy of ASHC increases 2.7% averagely, and after adjusting tags, these metrics are improved more than 3.9%. Based on this, tag systems with higher precision and applicability can be built.