The ultimate goal of many Internet-related research activities is to “transform” the web into a user-friendly environment where users can easily find things they are looking for. In many cases, this translates into a task of finding information useful for a user in a quick and efficient way. In the growing amount of information available on the Internet this errand becomes a challenge. Systems targeting information retrieval activities on the web are being equipped with user profiling. User profiles represent users’ interests, and are used as a means to support extraction of relevant information. Continuous updating of profiles following users’ activities on the web is a very important aspect of the profiling.
　　In this paper, we propose a new method for updating a user profile. The method is based on analyzing user’s browsing behavior, and identifying the most relevant items that should be added to the profile. This process uses a semantic-based similarity measure. This measure is estimated using rules representing multiple facets of similarity. The rules are constructed using a domain ontology. Obtained similarities are aggregated using Ordered Weighted Averaging (OWA) operator. This allows for expressing different levels of “strictness” in estimating similarity, and utilization of linguistic quantifiers OR, SOME, and MOST for that purpose. The semantic similarity is then combined with the items’ importance measures in order to identify items that are of the highest relevance to the user interests. The proposed approach is used for updating a profile in the music domain. The details of the real-word experiment are described in the paper.