tag-based user model) could lead to simple, viable and efficient approach to user modeling for adaptive web-based system. The challenge is then to combine the user model with the models of communities coming from the emerging social web and create a solid platform for personalization based on both traditional(e. g as presented in 3), and social approaches, such as the one presented in [4 When considering tag-based user models in a(tag-based)Web 2.0 environ ment, we are facing the need to be able to compare various tags. We might need to compare user characteristics or even whole user models, represented by tags o find similarities between users, which could serve for model maintenance as well as for more complex tasks such as online community creation or recom- mending in a recommender system. We might also need to compare the domai items represented by tags (e. g, a web page)in order to evaluate the explicit implicit feedback 5 and update the user model appropriately In this paper, we propose a method for inferring various relationships be- tween tags, which allows for full-blown usage of the tag-based user model for personalization and adaptation on the Web The paper is structured as follows: In section 2 we explain our approach to finding relationships between tags. Section 3 presents data we acquired and results of experiments we performed. In section 4 we summarize the related works, which served as an inspiration for our algorithm for building hierarchies from folksonomies. Finally, we give conclusions 2 Finding Relationships between Tags Our approach to finding relationships between tags combines three distinct ap- 1. Deriving of parent-child relationships between tags from a given folksonomy 2. Determining similarity between tags by applying spreading activation on the top of the folksonomy graph 3. Interconnecting tags by additional semantic relationships as well as enriching the whole tag corpus by adding external keywords: both coming from the Wordnet lexical database 2.1 Building Hierarchies from Folksonomies lksonomy is defined as a hypergraph 6h: where the set of vertices V=AUTUI and AnT=0, AnI=0, TOI=O and the set of ternary edges E={(a,t,)|a∈A,t∈T,t∈}. A social tagging system can be represented by such a hypergraph with following definitions of the sets A, T and i (we will use them for the rest of the paper as well) Actors(users)A=a1,.,ak] Tags(keywords, concepts)T=ti, t) Items(objects, instances)I=i1,.,imItag-based user model) could lead to simple, viable and efficient approach to user modeling for adaptive web-based system. The challenge is then to combine the user model with the models of communities coming from the emerging social web and create a solid platform for personalization based on both traditional (e.g., as presented in [3]), and social approaches, such as the one presented in [4]. When considering tag-based user models in a (tag-based) Web 2.0 environ￾ment, we are facing the need to be able to compare various tags. We might need to compare user characteristics or even whole user models, represented by tags, to find similarities between users, which could serve for model maintenance as well as for more complex tasks such as online community creation or recom￾mending in a recommender system. We might also need to compare the domain items represented by tags (e.g., a web page) in order to evaluate the explicit or implicit feedback [5] and update the user model appropriately. In this paper, we propose a method for inferring various relationships be￾tween tags, which allows for full-blown usage of the tag-based user model for personalization and adaptation on the Web. The paper is structured as follows: In section 2 we explain our approach to finding relationships between tags. Section 3 presents data we acquired and results of experiments we performed. In section 4 we summarize the related works, which served as an inspiration for our algorithm for building hierarchies from folksonomies. Finally, we give conclusions. 2 Finding Relationships between Tags Our approach to finding relationships between tags combines three distinct ap￾proaches: 1. Deriving of parent-child relationships between tags from a given folksonomy; 2. Determining similarity between tags by applying spreading activation on the top of the folksonomy graph; 3. Interconnecting tags by additional semantic relationships as well as enriching the whole tag corpus by adding external keywords; both coming from the Wordnet lexical database. 2.1 Building Hierarchies from Folksonomies Folksonomy is defined as a hypergraph [6] H := hV, Ei, where the set of vertices V = A ∪T ∪I and A ∩T = ∅, A ∩I = ∅, T ∩I = ∅ and the set of ternary edges E = {(a, t, i) | a ∈ A, t ∈ T, i ∈ I}. A social tagging system can be represented by such a hypergraph with following definitions of the sets A, T and I (we will use them for the rest of the paper as well): – Actors (users) A = {a1, ..., ak} – Tags (keywords, concepts) T = {t1, ..., tl} – Items (objects, instances) I = {i1, ..., im}