Akademik Veri Yönetim Sistemi
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), cilt.1892, ss.319-322, 2000 (SCI-Expanded, Scopus)
Incremental learning approaches based on user search activities provide a means of building adaptive information retrieval systems. To develop more effective user-oriented learning techniques for the Web, we need to be able to identify a meaningful session unit from which we can learn. Without this, we run a high risk of grouping together activities that are unrelated or perhaps not from the same user. We are interested in detecting boundaries of sequences between related activities (sessions) that would group the activities for a learning purpose. Session boundaries, in Reuters transaction logs, were detected automatically. The generated boundaries were compared with human judgements. The comparison confirmed that a meaningful session threshold for establishing these session boundaries was confined to a 11-15 minute range. © Springer-Verlag Berlin Heidelberg 2000.