Integrating Semantic Web technologies and Multi-Agent Systems: a Semantic Description of Multi-Agent Organizations Alexandra-MȃdȃlinaZarafin University Politehnica of Bucharest
Romania
AntoineZimmermann École Nationale Supérieure des Mines FAYOL-ENSMSE LSTI
F-42023 Saint-Étienne France
OlivierBoissier École Nationale Supérieure des Mines FAYOL-ENSMSE LSTI
F-42023 Saint-Étienne France
Integrating Semantic Web technologies and Multi-Agent Systems: a Semantic Description of Multi-Agent Organizations 9F5AFD6D42F5A76D9F97BA67A6F93A13 GROBID - A machine learning software for extracting information from scholarly documents

Introduction. Decentralization and openness are inherent properties of multiagent systems (MAS). The technologies they provide are thus the right abstraction for developing Web-oriented applications. Moreover, different works have been proposed to use Semantic Web technologies (SWT) for representing various dimensions of MAS (e.g., interaction protocols, norms, organizations). Given these facts, we think it is time to go a step further by integrating SWT and MAS in order to improve reusability of data, knowledge, coordination strategies, etc. on the Web and across systems. In this paper, we take a first step in this direction by proposing a semantic description of multi-agent organizations, showing the benefit regarding integration with Web ontologies.

Position. The Web is pervasive, populated with data, services, people and things, connected to each other, based on Web Standards. In such an environment, knowledge and semantic technologies become a necessity and a reality. Indeed, the Semantic Web has seen an increasing use not only for Web applications but in many areas of computer sciences. It encourages uniformity of data formats, as well as modularization and reuse of specifications (ontologies), by making it possible for ontologies to include and refer to information provided by other ontologies. Consequently, recent research in MAS have seen an intensive use of Knowledge representation together with increasing use of SWT. We envision that SWT will ultimately play a central role in all parts of MAS. Thus far, work combining MAS and SWT have only been concerned about addressing one dimension of MAS at a time. Also, they were mostly tackling the agent [6,4] and interaction dimensions [10,1] to ease communications, especially on domain knowledge. Other works have used those technologies to model part of the organization structure [8], norms and commitment [3], reputation [7] and more.

The situation shows that it is time to go beyond these ad hoc solutions and integrate the pieces into a complete Semantic-Web-based infrastructure. We observe too that none of the mentioned contributions were really taking advantage of the Web aspect of these technologies, except some Web service integration. We want to provide the specifications as Web data as well, such that agents can uniformly query and reason about agent systems, Web services, data and ontologies.

Semantic Web technologies and Multi-Agent Organizations. We made a step in this direction, focusing on the organizational aspect of MAS, which has been little addressed, although human organizations are of prominent interest to the Semantic Web community [9]. We used OWL to make an ontology of the organizational specification of the MOISE model [5], taking analysis of [2] as a basis. A given organization is described in RDF by instantiating the classes of the main ontology. The specification is made available at a SPARQL query endpoint, provided by an organizational artifact, so that agents who need to enter the organization can ask for relevant information on, e.g., roles or goals. Interestingly, agents can combine this with other data sources on the Web, using the same query language and protocol. Moreover, the specification can be combined with the agent's knowledge, who can specify constraining axioms in OWL (e.g., agents with whom I'm working must be disjoint from my friends) to detect undesired situations with consistency checks. Following Semantic Web good practices, we designed the ontology modularly and made it available online3 . To allow compatibility with the existing implementation of MOISE, we implemented a two-way transformation.

Conclusion.

MAS are increasingly interacting with the Web and we expect this to require the use of (1) Web standards and (2) semantic technologies. Therefore we want to encourage the use of SWT to enable the next generation of MAS. We contributed to this overall goal by providing an ontology for organization specifications.

http://purl.org/NET/maorg/moise
OWL-DL as a FIPA-ACL content language Proc. of FOCA 2006 of FOCA 2006 2006 Modelling Dimensions for Agent Organizations LCoutinho JSichman OBoissier Handbook of Research on Multi-Agent Systems: Semantics and Dynamics of Organizational Models 2009 Representation and monitoring of commitments and norms using OWL NFornara MColombetti AI Comm 23 4 2010 MHadzic EChang PWongthongtham Ontology-Based Multi-Agent Systems Springer 2009 A Model for the Structural, Functional, and Deontic Specification of Organizations in Multiagent Systems JHübner JSichman OBoissier Proc. of SBIA 2002 of SBIA 2002 2002 JASDL: A Practical Programming Approach Combining Agent and Semantic Web Technologies TKlapiscak RBordini Proc. of DALT 2008 of DALT 2008 2009 Experiments on semantic interoperability of agent reputation models using the SOARI architecture LNardin ABrandão JSichman Eng. Appl. of AI 24 8 2011 Designing MAS Organisation through an Integrated MDA/Ontology Approach DOkouya LPenserini SSaudrais AStaikopoulos VDignum SClarke Proc. of TWOMDE 2008 of TWOMDE 2008 2008 An organization ontology -W3C Working Draft 05 DReynolds April 2012. 2012 Working draft Agent Communication in DAML World YZou TFinin YPeng AJoshi RSCost Proc. of WRAC 2002 of WRAC 2002 2002