Using Semantic Relations for Representing Long-Term User Interests

In this paper we present a new mechanism for representing the longterm interests of a user in a user profile. Semantic relatedness between the profile terms is measured by using the web counting based method. Profile terms are associated through their sets of inductions words, representing highly related words to the terms that are found out through their co-occurrence in the web documents and semantic similarity. The relation between the two profile terms is then calculated using the combination of their corresponding sets of induction words. Although we have used the mechanism for long-term user profiling, applications can be more general. The method is evaluated against some benchmark methods and shows promising results.