A Label Quality-Oriented Method for Chinese Web Search Results Clustering

Abstract

Web search results clustering is an important strategy of organizing the snippets for modern search engine. Due to the semantic problems, there are still many challenges. This paper proposes a method named label quality-oriented scheme (LQOS) to achieve the clustering quality in the aspect of label. Specialized for Chinese queries, LQOS is designed according to the criterion of “readable label”, aiming at reducing noisy labels, and achieving the informative, complete and concise labels. Experiments on Sougou Chinese queries show that LQOS outperforms Lingo algorithm in the metrics of labels quality and the percentage of clustered snippets.