Intelligent Information Systems 9999 ISBN 666-666-666, pages 1–14 Personalized Search in Folksonomies with Ontological User Profiles Noriko Tomuro and Andriy Shepitsen College of Computing and Digital Media DePaul University Chicago, Illinois, USA Abstract This paper presents a new method for search personalization in Folksonomies which utilizes ontological user profiles. Notably, our method builds a folksonomy tag ontology in which the tags are disambiguated – each node corresponds to a single concept, and ambiguous tags are mapped to several nodes in the ontology. Our method first creates a set of unambiguous tag clusters by using an algorithm called DSCBC. Then we use a modified hierarchical agglomerative clustering al- gorithm to construct a disambiguated tag ontology. Next we match the tag profile of a target user against the ontology and derive an ontological profile of the user. Finally we feed the user’s ontological profile into the modified FolkRank algorithm and retrieve web resources which are ranked and personalized to the user. We ran our system on data from two social tagging systems. The results showed our method achieved significant improvements over other approaches. Keywords: Search Personalization, Folksonomies, Clustering, Natural Language Processing, FolkRank, Ontological User Profile 1 Introduction Collaborative tagging systems, usually referred to as folksonomies, is a popular new trend where Internet users apply descriptive tags to online resources. Central to collaborative tagging are annotations - users assign personalized tags to Web resources. The last few years has shown a rapid growth of various folksonomies. Two popular folksonomies are Delicious 1 and Last.FM 2 . In Delicious, users book- mark Universal Resource Locators (URLs). Last.FM allows users to upload, share and tag media files. Folksonomies are found in other applications such as on-line digital pictures, blogs and journal publications. The content of some of these re- sources can be determined by computers and effectively processed by standard search techniques, but some resources are particularly difficult to automatically categorize. For example, it can be particularly difficult for standard search engines 1 delicious.com/ 2 www.last.fm