Semantic web services represent an important and actual research area in computer science. A very popular topic in this area is the composition of semantic web services, which can be used for obtaining new semantic web services from existing ones. Based on a representation method for the semantic descriptions of semantic web services, that we had previously proposed, we propose a multi-agent system for the composition of semantic web services based on complexity functions and learning algorithms. Our system starts as a semi-automatic composition system, but after it gathers (using learning algorithms) sufficient information about the knowledge domain in which it is used, the system is able to perform compositions of semantic web services automatically. Based on the previously proposed representation method, this paper describes the structure and the main algorithms of the proposed system. The paper also presents an example of using the proposed system and some experimental results.Keywords: Semantic Web Service, Composition Of Semantic Web Services, Multi-Agent System, Complexity Functions, Learning Algorithms(ProQuest: ... denotes formulae omitted.)IntroductionThese days, semantic web services represent an important and actual research area in computer science. One of the main topics in this area is the composition of semantic web services, the main goal being to obtain new semantic web services by composing existing ones.In [1] the we propose the following classes of semantic web services composition methods: semi-automatic composition (see [2]), AI planning (see [3][4][5]), agents and multiagent systems (see [6][7]), logic languages and rules (see [8][9]), and bio-inspired methods (see [10][11]). One of the main conclusions of our analysis presented in [1] is that most of the semantic web services composition methods are automatic methods.In [12] we propose a method for representing the semantic description of a semantic web service using complexity functions; for information related to complexity functions, see, for example [13][14]. This new representation method is presented in [12] in a formal way, by proposing several definitions and theorems.In this paper we propose a multi-agent system for the composition of semantic web services based on the semantic descriptions representation method proposed in [12]. At the beginning, our method is semi-automatic, since it gathers information about the knowledge domain in which it is used. Later, after several uses of the method for the same knowledge domain, the method becomes automatic and it doesn't ask new information from the user. The main idea is that our method starts by being semi-automatic, learns the knowledge domain in which it works by using several learning algorithms and then it becomes automatic and doesn't need new information from the user for solving the semantic web services composition problem. Consequently, given a knowledge domain to work with, and a period of training, our method can be considered an automatic composition method.Taking into account the analysis made in [1], we can say that our automatic method belongs to the class 'agents and multi-agent systems'. An important element of originality of our composition method is that it uses a new representation method for the semantic descriptions, the one proposed in [12].The paper is organized as follows. Section 2 presents some concepts proposed in [12] that are necessary for this paper. Section 3 proposes a way to enrich the semantic description of a semantic web service (represented using the method proposed in [12]). Section 4 describes the proposed multi- agent system for the composition of semantic web services. In Section 5 we present the main algorithms used by our system. Section 6 contains an example of using the proposed system. In Section 7 we present some experimental results. Finally, Section 8 contains the conclusions of the paper. …
Read full abstract