Semantic User Model Inferences for Travel Recommender Systems

Introduction

Personalization offers many opportunities to improve the way web information is delivered to the users with different culture and backgrounds. Among personalization tools developed so far, recommender systems and Web personalization are successful examples of solutions recently introduced to improve Web documents searching, and information filtering such as products recommendation in E-Commerce (Shahabi & Chen, 2003).

Electronic tourism is one of the leading industries of E-business. In the travel and tourism domain many systems have been implemented to support personalized services. A wide range of heterogeneous information is available, and the complexity of product descriptions in the field of tourism is growing (Werthner & Klein, 1999). Travel Recommender Systems (TRS) are viewed as one of the important application domains, in order to assist the user in traveling, e.g., in an urban space. They are intuitive and valuable extensions to, and meanwhile a common means for tourism information systems based on observations in the real world (Berka & Plößnig, 2004).

In most user modeling and preference elicitation applications, there are many cases where no sufficient information and assumptions about the user are available to support user preference elicitation and personalization strategies. Usually, the user may either distrust a personalization system or be reluctant to be simulated and tracked by a user modeling component due to some privacy issues. In case of being required to fill a registration form, the user may do it so quickly that some inevitably incomplete user profiles and inconsistency of user’s information occur. On the other hand, concerning user modeling in the tourism domain, there is still to explore semantic user model and inference rules for travel recommender systems, especially with consideration of spatial criteria. Amongst many techniques currently developed, the semantic web provides one of the most promising solutions for describing and inferring user models, deriving inference rules for approximating user preferences. Semantic user models play a kernel role for recommender and web personalization systems. In a knowledge-based system, a semantic user model serves as an important input for personalizing human-computer interactions and interfaces. Particularly, travel recommender systems have direction relationships with the spatial dimension. That is, the spatial criteria, relationships and cognition have big infulences on the perceptions and interactions between and the user and information systems. However, to the best of our knowledge, there is few (if any) research to explore the effects of the spatial dimension on the representation of user model and the generation of recommender services.

This chapter introduces a semantic user model and inference rules based on a description logic language to represent and manipulate user information relevant to Travel Recommender Systems, with particular attention to the spatial domain. Description logics are effective to describe user’s information and knowledge at the semantic level, besides, they can efficiently handle inconsistency checking and incompletion issues, e.g., infer missing user information items. The logic-based modelling framework acts as a support for user classification and semantic web personalization, to favour the generation of additional information services to the users. The semantic user model can be used to represent a user’s knowledge and information (i.e. demographics), and with a set of domain-dependent rules specific to the tourism domain in terms of spatial criteria (i.e. distance) and cognition to infer useful user features such as interests and preferences. The user features serve as important inputs for Travel Recommender Systems to generate and deliver semantically rich, personalized information services to users. The objective of this research is to enhance the representation, reasoning and inference of user’s information and knowledge, such as interests and preferences, with semantic techniques and taking spatial cognition and criteria into account. The user-modelling approach is illustrated in the context of a tourism case study.

The remainder of this chapter is organized as follows. The second section, “User modelling and preference elicitation,” presents a semantic user model and inference rules for travel recommender systems. The section “DL user model inferences: an illustrative case” gives an illustrative case in the e-tourism domain. In the fourth section we discuss some implementation issues, and conclude this chapter in the last section.