Building User Groups Based on a Structural Representation of User Search Sessions

Identifying user groups is an important task in order to personalise search results. In Digital Libraries, visited resources and the sequential search patterns are often used to measure user similarity. Whereas visited resources help to understand what users want, they do not reveal how users prefer to search. In contrast, sequential patterns allow to decode the way in which users search, but they are very strict and do not allow changes in the order of the search. A third alternative and compromise could be the analysis of the structure of a search session. In this paper, we aim to obtain some insights into the potential of analysing search sessions on a structural basis. Therefore, we will investigate a structural representation of search sessions based on tree graphs. We will present a novel method to merge multiple session trees into a combined tree. Based on combined tree taken from similar sessions, we will build archetypical trees for different user groups.