Semantic-Based Service Recommendation ...

Semantic-Based Service Recommendation Framework for UCWW Haiyang Zhang, Nikola S. Nikolov, Ivan Ganchev Telecommunications Research Centre (TRC), University of Limerick, Ireland E-mail: {Haiyang.Zhang, Nikola.Nikolov, Ivan.Ganchev}@ul.ie Keywords: Electronic and Computer Engineering; Computer Science and Information Abstract Context-aware recommendation systems make recommendations by adapting to user's specific situation, and thus by exploring both the user preferences and the environment. In this paper, the design of a context-aware service recommendation framework utilising semantic knowledge in the Ubiquitous Consumer Wireless World (UCWW) is outlined. The main objective of the framework is to point to users the 'best' service instances that match their dynamic, contextualised and personalised requirements and expectations, thereby aligning to the always best connected and best served (ABC&S) paradigm.

1. Introduction The Ubiquitous Consumer Wireless World (UCWW) is a significant change to the global wireless environment, setting out a generic consumer-centric techno-business model foundation for future wireless communications technologies and networks [1]. In UCWW, the consumer-user potentially can use the mobile service of any service provider (xSP) through the (available) access network of any access network provider (ANP). In this paper, we propose a semanticbased recommendation model organizing services and their related attributes in a heterogeneous network, named a UCWW heterogeneous service network (HSN). Based on the dynamically formed HSN, the service recommendation framework will facilitate the automatic discovery of the 'best' service instances available for use through the 'best' access network (provider), following the user-driven 'always best connected and best served (ABC&S) paradigm [1].

2. Related work Systems that provide users with predictions and recommendations of items by mining the relationship between users and items are known as recommendation systems (RS). They have become a significant research area both in industry and academia over the last decades. In recent years, both common-sense and domain-specific knowledge has become popular for incorporation into the so-called semantic RSs. Semantic RSs are characterised by making use of semantic web technologies to improve the quality of recommendations. Most of them employ a conceptbased method to enhance the effectiveness of mere text mining by using more informative features for contentbased approach during user/item profiling and matching. Yu et al. use PathSim measurement to combine implicit feedback with different types of entity

relationships in order to generate recommendations. They diffuse the observed feedback along different meta-paths to generate possible recommendation candidate, and use a matrix factorisation technique to calculate latent representation for users and items under semantic meaning accordingly[2].

3. Research methodology By taking advantage of context-aware recommendation approaches and semantic Web technologies, the proposed service recommendation framework first collects and extracts service information, and models the HSN dynamically, based on a given network schema. Then, profile kernels, referring to the minimal set of features describing the user preferences, are then extracted to model the user profile. A recommendation engine considering both the user profile and the current context (user- and network context) are applied to recommend ‘best’ service instances to users. The proposed architectural framework is depicted in Figure 1.

Figure 1. The proposed service recommendation framework, incorporating an interface to Google Places and a third-party authentication, authorisation and accounting (3P-AAA) platform.

4. References [1] M. O'Droma and I. Ganchev, "The creation of a ubiquitous consumer wireless world through strategic ITU-T standardization," Communications Magazine, IEEE, vol. 48, pp. 158-165, 2010. [2] X. Yu, X. Ren, Y. Sun, B. Sturt, U. Khandelwal, Q. Gu, et al., "Recommendation in heterogeneous information networks with implicit user feedback," presented at the Proceedings of the 7th ACM conference on Recommender systems, Hong Kong, China, 2013.

Semantic-Based Service Recommendation Framework for UCWW Haiyang Zhang, Nikola S. Nikolov , Ivan Ganchev Telecommunications Research Centre (TRC), University of Limerick . Introduction

Aim

User modelling

The Ubiquitous Consumer Wireless World (UCWW) is a significant change to the global wireless environment, setting out a generic consumer-centric techno-business model foundation for future wireless communications technologies and networks [1]. In UCWW, the consumer-user potentially can use the mobile service of any service provider (xSP) through the (available) access network of any access network provider (ANP). In this paper, we propose a semantic-based recommendation model organizing services and their related attributes in a heterogeneous network, named a UCWW heterogeneous service network (HSN). Based on the dynamically formed HSN, the service recommendation framework will facilitate the automatic discovery of the 'best' service instances available for use through the 'best' access network (provider), following the user-driven 'always best connected and best served (ABC&S) paradigm [1].

• The goal of this service recommendation system is to provide users with an ordered list of service alternatives by taking into account the personalized, contextualized and dynamic requirements and expectations of the users, which means that the rank of services in the recommendation list is not only based on the relevance to that particular user (user context), but it should also take into account the current constrains of the access network (network context) and the specific requirements of the service (service context). • A major challenge in such a context-aware recommendation system is that due to the dynamic nature of the context, target instances to recommend vary in type within a scalable heterogeneous information network. Motivated by this challenge, the main problem to be solved is how effectively to model users and services in a scalable and efficient solution for real-time UCWW service recommendations.

A user profile usually includes long-term interests in using particular services, while the user-related context information includes features such as location, time, emotion, weather, etc. In a heterogeneous network, the reason of user A for choosing service S is a meta path, i.e. a path of nodes of potentially different type [3]. We propose to model a user profile based on the intersection of meta paths. We call the intersection of all meta paths, which represent the reasoning of user, a profile kernel. The profile kernel would be the minimal set of features which describe the user preferences and can be used to represent a user profile. An efficient method for extracting and updating profile kernels in real time will be developed as part of this work.

Concerning various types of contextual information, we leverage both pre-filtering and post-filtering approaches [4] depending on specific contextual attributes. Google Places API [5] will be employed as a geographical resource to guarantee that only services within certain distance are considered. For this data we will use a pre-filtering approach which will significantly reduce the data scale for a conventional 2D recommendation stage. A refined hybrid recommendation method which relies on a graph-based similarity measurement will be elaborated and implemented to produce the final recommendation list. This method should be able to recommend services along with three criteria: service quality, user specificity, and diversity [8].

Method By taking advantage of context-aware recommendation approaches and semantic Web technologies, the proposed service recommendation framework first collects and extracts service information, and models the HSN dynamically, based on a given network schema. Then, profile kernels, referring to the minimal set of features describing the user preferences, are then extracted to model the user profile. A recommendation engine considering both the user profile and the current context (user- and network context) are applied to recommend ‘best’ service instances to users. The proposed architectural framework is depicted in Figure 1. Figure 1 The proposed service recommendation framework, incorporating an interface to Google Places and a third-party authentication, authorization and accounting (3P-AAA) platform).

Heterogeneous service network Due to privacy restrictions, the amount of user information available to exploit and analyses is very limited, while services are usually fully described by service providers (xSPs). Unlike most recommendation systems which focus on user modeling, the UCWW RS will use also detailed service features exploited from the service model. Mobile services are described by means of service descriptions (SDs), which are categorized in two classes ANP SDs and xSPs SDs [2]. ANP SDs are used by the user to find and use the best' access network available in the current location, while xSPs SDs are more complex. They are based on a set of context parameters, classified in three groups -user-, service-, and network context.

Recommendation Engine

Conclusion The proposed system will be employed for discovering the 'best' service instances available for use through the 'best' access network (provider), realizing the user-centric ABC&S paradigm [1]. Future will seek further specification of the system design, followed by implementation, testing and evaluation of a system prototype in accordance with the steps outlined.

Reference [1] M. O'Droma and I. Ganchev, "The creation of a ubiquitous consumer wireless world through strategic ITU-T standardization," Communications Magazine, IEEE, vol. 48, pp. 158-165, 2010. [2] Z. Ji, I. Ganchev, and M. O’Droma, "Advertisement Data Management and Application Design in WBCs," Journal of Software, vol. 6, pp. 1001-1008, 2011 Figure 2. A heterogeneous service network example, and a hypothetic real world eample.

[3] Y. Sun, J. Han, X. Yan, P. S. Yu, and T. Wu, "Pathsim: Meta path-based top-k similarity search in heterogeneous information networks," VLDB’11, 2011. [4] G. Adomavicius and A. Tuzhilin, "Contextaware recommender systems," in Recommender systems handbook, ed: Springer, 2011, pp. 217-253. [5] Google Places API. https://developers.google.com/places/documentatio n/.