Implication of tagging behavior on ontology construction

Abstract

The ontology is expected to solve many language problems in text mining, translation, and web search. It can be constructed by some experts in small scale, but it is not a scalable solution. Query log analysis (QLA) is an alternative solution to construct ontology. Through mining query logs, a basic ontology can be constructed automatically. However, QLA has some problems: not public, hard to evaluate, and ambiguous to understand context. To solve these problems, we propose to use tagging information to construct ontology automatically. Tagging information may have QLA`s strong points and also overcome QLA`s problems. In this paper, we quantitatively show that tagging information is comparable with query logs in ontology construction. Specifically, we investigate user behavior and find a tagging pattern to construct the ontology.