Logo des Repositoriums

Auflistung nach Autor:in "Jannach, Dietmar"

1 - 4 von 4
  • Zeitschriftenartikel
    Explaining Online Recommendations Using Personalized Tag Clouds
    (i-com: Vol. 10, No. 1, 2011) Gedikli, Fatih; Ge, Mouzhi; Jannach, Dietmar
    Recommender systems are sales-supporting applications that are usually integrated into online shops and are designed to point the visitor to products or services she or he might be interested in but has not bought yet. In the last decade, many techniques have been developed to improve the predictive accuracy of such systems. However, there are also factors other than accuracy that infl uence the user-perceived quality of such a system. In particular, system-generated explanations as to why a certain item has been recommended have shown to be a valuable tool to improve both the user's satisfaction and the system's effi ciency. This paper reports the results of a fi rst user study which was conducted to evaluate whether personalized tag clouds are an appropriate means to visually explain recommendations. The evaluation reveals that using tag clouds as explanation mechanism leads to higher user satisfaction and recommendation effi ciency than previous keyword-style explanations.
  • Konferenzbeitrag
    Explanations and User Control in Recommender Systems
    (Proceedings of the 23rd International Workshop on Personalization and Recommendation on the Web and Beyond, 2019) Jannach, Dietmar; Jugovac, Michael; Nunes, Ingrid
  • Konferenzbeitrag
    Item Familiarity as a Possible Confounding Factor in User-Centric Recommender Systems Evaluation
    (i-com: Vol. 14, No. 1, 2015) Jannach, Dietmar; Lerche, Lukas; Jugovac, Michael
    User studies play an important role in academic research in the field of recommender systems as they allow us to assess quality factors other than the predictive accuracy of the underlying algorithms. User satisfaction is one such factor that is often evaluated in laboratory settings and in many experimental designs one task of the participants is to assess the suitability of the system-generated recommendations. The effort required by the user to make such an assessment can, however, depend on the user’s familiarity with the presented items and directly impact on the reported user satisfaction. In this paper, we report the results of a preliminary recommender systems user study using Mechanical Turk, which indicates that item familiarity is strongly correlated with overall satisfaction.
  • Zeitschriftenartikel
    Perspektiven in der Offline-Evaluation von Empfehlungsalgorithmen
    (HMD Praxis der Wirtschaftsinformatik: Vol. 50, No. 5, 2013) Jannach, Dietmar; Lerche, Lukas
    enEmpfehlungssysteme sind heutzutage ein zentraler Bestandteil vieler Onlineshops und stellen für die Betreiber ein wertvolles Mittel dar, Kunden bei der Produktoder Informationssuche zu helfen sowie auf weitere interessante Produkte hinzuweisen. Die meisten Forschungsarbeiten zu Empfehlungssystemen verwenden explizite Produktbewertungen von Kunden als Eingabe für die Algorithmen und als Grundlage für die Empfehlungsgenerierung. In der Realität sind solche Bewertungen jedoch oft nicht in ausreichender Menge vorhanden, sodass für die Produktvorschläge auf andere Datenquellen — wie zum Beispiel Logdaten der Kundenaktionen — zurückgegriffen werden muss. In diesem Beitrag werden praktische Herausforderungen bei der Nutzung und Interpretation solcher weiteren Datenquellen für die Empfehlungsgenerierung besprochen sowie auf methodische Fragen der vergleichenden Bewertung von Empfehlungsalgorithmen eingegangen.