A semantic enhanced hybrid recommendation approach: A case study of e-Government tourism service recommendation system

Highlights

• A hybrid semantic enhanced recommendation approach

• A new Inferential Ontology-based Semantic Similarity (IOBSS) between two ontological instances

• A few new concepts: Association, Associate Network and Common Associate Pair Set

• A case study of Australian e-Government tourism services

Abstract

Recommender systems are effectively used as a personalized information filtering technology to automatically predict and identify a set of interesting items on behalf of users according to their personal needs and preferences. Collaborative Filtering (CF) approach is commonly used in the context of recommender systems; however, obtaining better prediction accuracy and overcoming the main limitations of the standard CF recommendation algorithms, such as sparsity and cold-start item problems, remain a significant challenge. Recent developments in personalization and recommendation techniques support the use of semantic enhanced hybrid recommender systems, which incorporate ontology-based semantic similarity measure with other recommendation approaches to improve the quality of recommendations. Consequently, this paper presents the effectiveness of utilizing semantic knowledge of items to enhance the recommendation quality. It proposes a new Inferential Ontology-based Semantic Similarity (IOBSS) measure to evaluate semantic similarity between items in a specific domain of interest by taking into account their explicit hierarchical relationships, shared attributes and implicit relationships. The paper further proposes a hybrid semantic enhanced recommendation approach by combining the new IOBSS measure and the standard item-based CF approach. A set of experiments with promising results validates the effectiveness of the proposed hybrid approach, using a case study of the Australian e-Government tourism services.

Introduction

Recommendation systems (RSs) are known as the most popular applications of Web personalization. The RSs aim to provide users with personalized services or products that are relevant to their needs and interests. Recent research studies show that existing personalized online services adopt several RSs approaches. These approaches are classified into four main categories, including content-based (CB) filtering, collaborative filtering, knowledge-based filtering and hybrid recommendation [1], [10], [40]. Although the CB filtering and CF approaches are the most popular in practical applications, both of them suffer from several limitations [23]. For instance, the CB filtering approach tends to result in overspecialization in which the diversity in the recommendation results eventually vanishes [35], while the CF approach suffers from the data sparsity problem which occurs when the ratings obtained are few compared to the number of available items. Moreover, both the CB filtering and CF approaches have difficulty offering accurate recommendations for new items as there is usually little available information about new items.

On the other hand, hybrid recommendation approaches, as a combination of two or more recommendation approaches, have been proposed to overcome the main limitations of traditional recommendation approaches and improve the quality of the recommendations offered [1], [11], [35]. Most of the existing hybrid recommendation approaches combine conventional CF approaches with other approaches such as CB filtering, since CF approaches are generally known to be the most promising approaches in the recommendation systems domain [1], [23], [45]. There has been considerable research into the hybridization of CF-based algorithms and improvements on the prediction accuracy have been made [11], [12], [45], [50]. However, obtaining better prediction accuracy and overcoming the main limitations of the standard CF recommendation approaches remain open challenges, as no cure-all solution is yet available and many research studies have been working on solutions for each of the CF limitations [12], [45].

These challenges, combined with the increasing popularity of semantic web technologies, have inspired a growing interest in semantic enhanced recommendation approaches. These approaches mainly incorporate the semantic knowledge of users and/or items within the recommendation process of conventional CF-based algorithms to accurately evaluate similarity of items and to enhance recommendation accuracy [8], [36]. Most of these approaches rely on semantic knowledge extracted from a target ontology that includes the direct hierarchical (i.e. taxonomical) relationships of items and/or their shared attributes. However, evaluating the similarity of items is limited since ontological relationships¹ that connect the items in a target ontology are not usually handled very well [7], [25], [26], [33], [44]. Such relationships may include complex relationships between instances (i.e. items²) that consist of two or more relationships [3].

Even though progress is being made in developing efficient strategies for estimating the semantic similarity of items in semantic enhanced recommendation systems, this work is still in an early stage and more research is needed [3], [8], [13], [15], [25], [44]. This observation, combined with the specific features of service items (e.g. services are multi-relation and highly interrelated) in a specific domain, such as services in government, has motivated the research presented in this paper. Consequently, this paper presents two contributions (i) it proposes a new IOBSS measure to evaluate the semantic similarity between instances in specific domain ontology and (ii) it develops a new semantic enhanced hybrid recommendation approach that combines the new semantic similarity measure and the item-based CF to generate accurate recommendations.

The effectiveness of the new semantic-based hybrid recommendation approach has been validated through a case study of the Australian e-Government tourism service. It achieves highly effective results in terms of prediction accuracy of generated recommendations and in alleviating data sparsity and cold-start new item problems.

The rest of the paper is organized as follows. Section 2 presents the related work. Section 3 presents the concept and calculation procedure of the new IOBSS measure with an illustrative example. Section 4 presents the new semantic-based enhanced hybrid recommendation approach, its workflow and its computation recommendation procedure. An experimental study of the new hybrid recommendation approach, in the context of recommending e-Government tourism services, is illustrated in Section 5. Finally, Section 6 concludes the paper and highlights potential future work.