Comparison and Evaluation of Highly Related Tag-Pairs Extraction Methods

Abstract

Until recently, it was expected that the web users' tagging would be used in classification, recommendation, and information search. However, the users' subjective decision or infrequent tags with high interrelationship have caused inaccurate results.This research proposes an algorithm to disregard users' subjective tags and utilize infrequent tags with high semantic similarity using WordNet. Furthermore, to enhance the efficiency of the proposed algorithm, a weighted matrix is proposed.To evaluate the proposed algorithm, the current co-appearing frequency between tagpairs method, tag-pair semantic similarity extraction algorithm, and tag-pair weight matrix method were analyzed and compared.

Design of Tag Ranking Algorithm Based on Cluster

“…A TWM(Tag-pair Weight Matrix) which has semantic similarity value between tag-pairs, was carried out in the preceding research using equation (1) [10].…”