On Deriving Tagsonomies: Keyword Relations Coming from crowd Michal barla and maria bielikova Institute of Informatics and Software Engineering, Faculty of Informatics and Information Technologies, Slovak University of Technology Ilkovicova 3. 842 16 Bratislava, Slovakia I Name Surname fiit. stuba sk Abstract. Many keyword-based approaches to text classification, in formation retrieval or even user modeling for adaptive web-based system could benefit from knowledge on relations between various key words which gives further possibilities to compare them, evaluate their distance etc. This paper proposes an approach how to determine keyword rela- tions(mainly a parent-child relationship) by leveraging collective wisdom of the masses, present in data of collaborative(social)tagging systems on the Web. The feasibility of our approach is demonstrated on the data coming from the social bookmarking systems delicious and CiteULik 1 Introduction Phenomenon of the Social Web, with its roots in Web 2.0, is gaining a lot of attention all over the world in both research and practice. We are studying the power and wisdom of masses, when millions of people switched from the pas- sive reading of the content to the active participation in its creation. People are blogging, sharing wikis, connecting themselves in various social applications nd above all-tagging almost everything they get into touch: bookmarks, pho- tographs, videos, publications, blogposts, articles etc. People got used to classify items by ng few simple tags to it and to use those tags for a future retrieval of their favorite items. More, they are often expecting to find a new, yet unseen content by using their own tags. a part of the Web 2.0 success lies in an implicit agreement of masses on a shared(but never explicitly defined) vocabulary used to tag items -folksonomies. At the same time, vast amount of content requires efficient navigation sup- port, content reorganization or filtering -personalization and adaptation of the web. Its efficiency is dependent on adaptive systems ability to capture and maintain user model. A lot of research was devoted to finding the most suitable fexible or most generic and all-encompassing user model representation [1, 2 however, so far we are not aware of any explicit agreement on an ideal model The obvious question, when analyzing the success of Web 2.0, is whether an assignments of keywords(tags)to user instead of to pages (i.e, creation ofOn Deriving Tagsonomies: Keyword Relations Coming from Crowd Michal Barla and M´aria Bielikov´a Institute of Informatics and Software Engineering, Faculty of Informatics and Information Technologies, Slovak University of Technology Ilkoviˇcova 3, 842 16 Bratislava, Slovakia {Name.Surname}@fiit.stuba.sk Abstract. Many keyword-based approaches to text classification, in￾formation retrieval or even user modeling for adaptive web-based system could benefit from knowledge on relations between various keywords, which gives further possibilities to compare them, evaluate their distance etc. This paper proposes an approach how to determine keyword rela￾tions (mainly a parent-child relationship) by leveraging collective wisdom of the masses, present in data of collaborative (social) tagging systems on the Web. The feasibility of our approach is demonstrated on the data coming from the social bookmarking systems delicious and CiteULike. 1 Introduction Phenomenon of the Social Web, with its roots in Web 2.0, is gaining a lot of attention all over the world in both research and practice. We are studying the power and wisdom of masses, when millions of people switched from the pas￾sive reading of the content to the active participation in its creation. People are blogging, sharing wikis, connecting themselves in various social applications and above all – tagging almost everything they get into touch: bookmarks, pho￾tographs, videos, publications, blogposts, articles etc. People got used to classify items by assigning few simple tags to it and to use those tags for a future retrieval of their favorite items. More, they are often expecting to find a new, yet unseen content by using their own tags. A part of the Web 2.0 success lies in an implicit agreement of masses on a shared (but never explicitly defined) vocabulary used to tag items – folksonomies. At the same time, vast amount of content requires efficient navigation sup￾port, content reorganization or filtering – personalization and adaptation of the web. Its efficiency is dependent on adaptive system’s ability to capture and maintain user model. A lot of research was devoted to finding the most suitable, flexible or most generic and all-encompassing user model representation [1, 2], however, so far we are not aware of any explicit agreement on an ideal model representation. The obvious question, when analyzing the success of Web 2.0, is whether an assignments of keywords (tags) to user instead of to pages (i.e., creation of