Mobile Search in a Preference World
Wolf‐Tilo Balke · 2002
INTRODUCTION We investigate the importance of focused search for mobile web- based services. Focused search aims at delivering only those query results to a mobile client that are best matches wrt. the cur- rent location, situation and individual preferences of the mobile user. Our approach is characterized by an intuitive and powerful notion of preferences, which are modeled as strict partial orders under a 'I like A better than B' semantics. The underlying theory of preferences, their implementation and integration into data- bases and Internet services has been investigated for several years within the research program Preference World at the University of Augsburg. For the scope of this paper we discuss some of its results with special attention to the topic of mobile search and we will outline ongoing research work. In particular we demonstrate how to do preference engineering for focused search applying the database query languages Preference SQL and Preference Xpath. The performance aspects of mobile services are addressed by architectural considerations and by a novel algorithm called SR- Combine for efficient top-k search. We show how techniques for progressive delivery, which change the delivery order preferring most relevant document parts within a given level of detail, can improve mobile search. Moreover, advanced personalization in mobile services will require to pay an even closer attention to the key notion of preferences, taking cognitive modeling aspects into account. In summary we claim that careful preference modeling and efficient preference-based queries are crucial for the success of mobile search and personalized mobile services. Appropriate focused search capabilities are an essential problem when designing mobile services. Due to low bandwidths and small display sizes, it is reasonable to return only a few result objects. However, these should have a high precision, because discarding irrelevant result objects using mobile devices is a very tedious task. To avoid the flooding of users with irrelevant ob- jects, the paradigm of top-k querying has proven to be a valuable feature in today's applications. Users specify the number k of objects to return and get only the k best matching objects from the application. However, this requires a cooperative answer behav- ior (1) delivering the best matching objects only, even if no exact matching objects were found. Thus also the well-known 'empty result effect' can be avoided.