A customized semantic service retrieval methodology for the digital ecosystems environment

Abstract

With the emergence of the Web and its pervasive intrusion on individuals, organizations, businesses etc., people now realize that they are living in a digital environment analogous to the ecological ecosystem. Consequently, no individual or organization can ignore the huge impact of the Web on social well-being, growth and prosperity, or the changes that it has brought about to the world economy, transforming it from a self-contained, isolated, and static environment to an open, connected, dynamic environment. Recently, the European Union initiated a research vision in relation to this ubiquitous digital environment, known as Digital (Business) Ecosystems. In the Digital Ecosystems environment, there exist ubiquitous and heterogeneous species, and ubiquitous, heterogeneous, context-dependent and dynamic services provided or requested by species. Nevertheless, existing commercial search engines lack sufficient semantic supports, which cannot be employed to disambiguate user queries and cannot provide trustworthy and reliable service retrieval. Furthermore, current semantic service retrieval research focuses on service retrieval in the Web service field, which cannot provide requested service retrieval functions that take into account the features of Digital Ecosystem services. Hence, in this thesis, we propose a customized semantic service retrieval methodology, enabling trustworthy and reliable service retrieval in the Digital Ecosystems environment, by considering the heterogeneous, context-dependent and dynamic nature of services and the heterogeneous and dynamic nature of service providers and service requesters in Digital Ecosystems.The customized semantic service retrieval methodology comprises: 1) a service information discovery, annotation and classification methodology; 2) a service retrieval methodology; 3) a service concept recommendation methodology; 4) a quality of service (QoS) evaluation and service ranking methodology; and 5) a service domain knowledge updating, and service-provider-based Service Description Entity (SDE) metadata publishing, maintenance and classification methodology.The service information discovery, annotation and classification methodology is designed for discovering ubiquitous service information from the Web, annotating the discovered service information with ontology mark-up languages, and classifying the annotated service information by means of specific service domain knowledge, taking into account the heterogeneous and context-dependent nature of Digital Ecosystem services and the heterogeneous nature of service providers. The methodology is realized by the prototype of a Semantic Crawler, the aim of which is to discover service advertisements and service provider profiles from webpages, and annotating the information with service domain ontologies.The service retrieval methodology enables service requesters to precisely retrieve the annotated service information, taking into account the heterogeneous nature of Digital Ecosystem service requesters. The methodology is presented by the prototype of a Service Search Engine. Since service requesters can be divided according to the group which has relevant knowledge with regard to their service requests, and the group which does not have relevant knowledge with regard to their service requests, we respectively provide two different service retrieval modules. The module for the first group enables service requesters to directly retrieve service information by querying its attributes. The module for the second group enables service requesters to interact with the search engine to denote their queries by means of service domain knowledge, and then retrieve service information based on the denoted queries.The service concept recommendation methodology concerns the issue of incomplete or incorrect queries. The methodology enables the search engine to recommend relevant concepts to service requesters, once they find that the service concepts eventually selected cannot be used to denote their service requests. We premise that there is some extent of overlap between the selected concepts and the concepts denoting service requests, as a result of the impact of service requesters’ understandings of service requests on the selected concepts by a series of human-computer interactions. Therefore, a semantic similarity model is designed that seeks semantically similar concepts based on selected concepts.The QoS evaluation and service ranking methodology is proposed to allow service requesters to evaluate the trustworthiness of a service advertisement and rank retrieved service advertisements based on their QoS values, taking into account the contextdependent nature of services in Digital Ecosystems. The core of this methodology is an extended CCCI (Correlation of Interaction, Correlation of Criterion, Clarity of Criterion, and Importance of Criterion) metrics, which allows a service requester to evaluate the performance of a service provider in a service transaction based on QoS evaluation criteria in a specific service domain. The evaluation result is then incorporated with the previous results to produce the eventual QoS value of the service advertisement in a service domain. Service requesters can rank service advertisements by considering their QoS values under each criterion in a service domain.The methodology for service domain knowledge updating, service-provider-based SDE metadata publishing, maintenance, and classification is initiated to allow: 1) knowledge users to update service domain ontologies employed in the service retrieval methodology, taking into account the dynamic nature of services in Digital Ecosystems; and 2) service providers to update their service profiles and manually annotate their published service advertisements by means of service domain knowledge, taking into account the dynamic nature of service providers in Digital Ecosystems. The methodology for service domain knowledge updating is realized by a voting system for any proposals for changes in service domain knowledge, and by assigning different weights to the votes of domain experts and normal users.In order to validate the customized semantic service retrieval methodology, we build a prototype – a Customized Semantic Service Search Engine. Based on the prototype, we test the mathematical algorithms involved in the methodology by a simulation approach and validate the proposed functions of the methodology by a functional testing approach.