Personalized Faceted Navigation in Semantically Enriched Information Spaces Michal Tvaroˇ zek, M´ aria Bielikov´ a Institute of Informatics and Software Engineering, Faculty of Informatics and Information Technologies, Slovak University of Technology Ilkoviˇ cova 3, 842 16 Bratislava, Slovakia {tvarozek,bielik}@fiit.stuba.sk Abstract Existing information retrieval systems provide users with limited support for efficient navigation in large semanti- cally enriched information spaces. Several possible solu- tions were proposed, such as using faceted metadata search or semantic clusters of search results. We explore the pos- sibilities of using enhanced faceted navigation with support for personalization, collaboration and Semantic Web tech- nologies for (semantic) information retrieval. Furthermore, we propose the extension of faceted browsers with support for dynamic facet generation based on an automatically ac- quired user model, and evaluate the proposed ideas in mul- tiple domains – scientific publications, digital images and job offers. 1 Introduction The present Web along with many web-based resources comprise a unique ubiquitous source of information and an environment for collaboration and interaction of many users and businesses. While the amount of available information and the quality and capabilities of information search and processing tools are growing at an incredible rate, so do the size and diversity of the Web’s user base and the expecta- tions and requirements of individual users. Although existing information retrieval (IR) methods are continuously improving, they still fail to address the in- creasing requirements and expectations of many users with specific needs. For example, most existing search engines such as Google or MSN Live Search employ keyword-based search, while sharing systems such as Flickr or YouTube might extend this with tag-based search. The infamous “ad- vanced search” interfaces allow users to specify even more complex (keyword-based) queries, optionally with some ad- ditional domain specific attributes (e.g., size, filetype for images). Video search sites such as IMDb and MovieLens take complexity to another level by offering (multistep) in- terfaces with many text fields, drop-down menus and multi- choice listboxes. However, several studies have repeatedly indicated that typical search queries are short (up to four words; depend- ing on the domain) [12] and that advanced search is im- practical to use for many users [21]. While existing sys- tems are generally good when searching for very specific items, they do not support browsing and exploratory tasks sufficiently [28]. A field study of journalists and news- paper editors selecting photos for newspaper articles con- ducted by Markulla and Sormunen reported that “profes- sional users” needed to search on multiple categories [16], yet found an elaborate advanced search interface with about 40 input forms unusable. The Web is a dynamic open information space as many “information artefacts” – documents, articles, images, videos, music files etc. are continuously added, modified, removed, rated or tagged. Thus, user diversity and the evo- lution of information and user characteristics over time play a crucial role in effective user-centred IR system design. For example, people who grew up with the Web and the Inter- net, i.e. the “Net Generation”, have a natural understanding of this new ubiquitous environment quite unlike their pre- decessors [18]. Consequently, they have (radically) new re- quirements, expectations and modes of operation compared to the previous generation of web users. Accordingly, current changes include a shift from tradi- tional lookup tasks (e.g., fact retrieval) towards more ad- vanced and open ended learning and investigation tasks (e.g., knowledge acquisition, comparison, aggregation, analysis or planning) collectively described as exploratory search [15]. Furthermore, the trend towards more interac- tion and active (social) participation encourages the combi- nation and cross-fertilization of approaches from human- computer interaction, information retrieval, the Adaptive Web and the Semantic Web. In this paper we build upon several existing approaches