Multi-agent pproach for Knowledge Processes between Research Organisations Kimmo Salmenjoki, Yaroslav Tsaruk, ki Vainio University of Vaasa, Finland Semantic knowledge and process oriented communication Knowledge systems for communicational process with agents Implementing a contextaware communication agent for a KMO KN portal http://www.w3.org/2001/sw/ 1
Modelling processes and information services Using extended semantic web concepts and related platforms and toolboxes we build ontology based, process oriented and dynamic knowledge descriptions model users needs and organisational processes in generic knowledge management algorithms Due to the usage of semantic web and ontologies this MS- CommonKDS model can be reduced in our case: develop jointly the gent model (generic), Task model (here for communication processes only), Communication model (between the agents) and the Design model Figure 1: MS-CommonKDS layers in design of agent based systems part of the KMO ontology the common knowledge processes in KN portal usage by individual persons first. The domain of communication in this figure consists of messages and then their atomic decomposition a part of the related research collaboration metadata, which will be used to create the context and metadata of the KMO communicational processes These processes will be set up to support both human and agent enhanced communication in vivo and between persons. Later this automation will be expanded for groups and communities of users, too Event KMO level R&D topic Org. activity ctivitylist CSE Topic ctionlist KMO action KMO node Research topic Data and activities Business processes Tasklist Task Topic Originator receiver Shared message Project Research project Organization Coordinator KMO node Communication Message tom message Figure 2: part of the higher ontology related to research collaboration 2
Organisational / personal processes in communication simple interaction patterns to advocate the personal, interpersonal and between-nodes communication. The dashed lines mediate between shared and individual views the goals for the agents in the communication processes describe the desired states of the system and its environment, where as tasks describe state transitions that are needed to satisfy agent goal commitments an automated actor is represented by the user supporting communication agent, but in more general cases by automated agents Research topic utomation KMO acts Collaboration & support Organization ctivity Personal context list Goals KMO ction Goals ction processes list Plan Information Shared tasks Operation Tasks Person User / gent acts Communication activity KMO Emails metadat a Information Figure 3: From email exchange to KMO communicational processes gents for other KMO modules (and knowledge processes) n action in our system consists of a planned and technically subdivided act. s shared operations and processes in collaboration any action pushes the act by allocating individual(s), timings, contexts and related coordinated actions and acts in the domain scope. n act in this figure consists of a real-world actor (or later agent) and the consumption of it s context. In principle, when the (communicational or other) ctions have been scheduled and timed, the (planned) acts become tasks gent model contains a detailed and comprehensive description of each individual agent, providing an internal view that includes the agent s goals and the services (or the functional capability) they provide. This contrasts and supplements the external perspective provided by the organization model as discussed in the previous section. lso these agent oriented concepts are equivalent to with the related information concepts of the ontology (partly presented in Figure 2) In fr a st ru ct ur e KMO collaboration and communities Semantic knowledge modules KMO node portals and search Organizational processes and services Semantic web toolbox platforms (like OWL with SWWS, WSML of Deri, GUN (with RscDF, RgbDF), KON, dvanced Knowledge Technologies KT, Haystack) KMO semantic content management KMO realtime collaboration KMO data analysis KMO R&D processes and services Semantic web infrastructure (Joseki, Jena, RDF, RDF-S and related technologies) Server infrastructure (web server, mail server, messaging services, web service and XML technologies) with Microsoft.NET 3
Capturing communicative knowledge in processes the external/internal approach described for agent software: One of the first methodologies for the development of Belief, Desire, Intension (BDI) agents (like mental attitudes) based on the object oriented (OO) technologies have been presented by Kinny, Geoff [6] and Kinny, Geoff, Rao [7]......Their agent methodology distinguishes between the external viewpoint- the system is decomposed into agents, modeled as complex objects characterized by their purpose, their responsibilities, the services they perform, the information they require and maintain, and their external interactions- and the internal viewpoint- the elements required by a particular agent architecture (an agent s belief, goals and plans) must be modeled for each agent. Figure 4: Overall agent structure for shared communication using agent architecture of MESSGE project KMO software agent roles CONTEXT_GENT COMM_GENT CONTENT_GENT CTOR_GENT the role of the CONTEXT_GENT role playing agents are to host the user context, and to provide the user simple, context-sensitive services on the information contained in the KN portals. t this stage, we want to test the viability of this kind of agent approach in the personal context only The CONTENT_GENT and CTOR_GENT types and their implementations will later provide the knowledge processing actions. The CONTENT_GENT is a group of automated agents that creates, supports and enhances the communication of content in KMO, based on the personal and group-wise interests and needs indicated by the CONTEXT_GENTS representing the individuals in the KMO collaboration The agents with the CTOR_GENT role are at higher level automated research processes within the local organisation (or their knowledge support), and later these will be sectioned and developed further to work between the collaborating KN node portals and persons in R&D Conversation EMailMessage SharedMessage PersonalMessa ge News Meeting Forum Message Content Include s CalendarSched ule Include s Calendar Event Figure 5: Communication part of the KMO ontology for the data used by the COMM_GENT 4
Email exchange patterns To cover the dynamic nature of the communication we will transform the static organizational concepts of KMO in Figure 2 into dynamic processes with procedural and algorithmic approaches traversing the tree the hierarchies Goal->ctivity -> ct -> ction -> Operation on the organizational branch represents both the organizational and agent approaches for establishing the KMO communication processes the iteration Goal->ctivity -> ction -> Operation of Figure 3 will be give our simplified COMM_GENT algorithm as following three agent actions -C triggered by the various events in the KMO communication Mail server Web portal Server - side components Extraction info from E-mail Representation infofmation from storage Storing information SW technologies with Jena and Joseki RDF storage COMM_GENT algorithms. ddition of a new resource (local only) (triggered by new event) 1. (default status:) ZOE is running for several users (with their email), KMO COMM_GENT is running on the TTWIN-server. t present stage the KMO COMM_GENT is emulated by a ZOE shared email server 2. When a shared message appear it will be added as resources of RDF statement into the system (later this should be done automatically by the COMM_GENT as it perceives the information entities) 3. COMM_GENT recognizes the new resource, and sees if it matches the KMO user s context. If so, it informs the related users, otherwise it adds the statements to the shared KMO knowledge base IDEs: combine semantic web and web service technologies towards classical knowledge management tasks. 5
or example scenarios B. Reacting to user context (triggered by user actions on COMM_GENT clients) 1. (default status:) Now ZOE later a KN portal client with JDE is running for several users (as a KMO context interface KMOCI client), KMO COMM_GENT is running on the TTWIN-server 2. When user searches for something in KMOCI, the COMM_GENT will activate the knowledge base (cumulated) and give semantic search results/ hints 3. Based on user actions/ preferences, COMM_GENT updates the personal user view of the KMO content (if necessary) IDE: various user focused services with this context, knowledge related tasks and automated processes according to some predefined scenarios and agendas... C. gent s internal action (default) 1. (default status:) KMOCI clients are running for several users (as a KMO context interface KMOCI), coordinating KMO COMM_GENT is running on the TTWIN-server 2. KMO COMM_GENT is running on each KN 3. COMM_GENTs maintain and share the KMO content in real-time and synchronized 4. (later, there will be other services, related to other KMO agents and processes like above, that the various agents coherently process) 2. ddition of a new resource (remote) (triggered by new event) 1. (default status:) ZOE is running for several users (with their email), KMO COMM_GENT is running on the TTWIN-server 2. When a shared message appear it will be added as resources of RDF statement into the system (locally) 3. COMM_GENT recognizes the new resource (from distance), and sees if it matches the KMO user s context. If so, it informs the related users, otherwise it adds the statements to the shared KMO knowledge base (and synchronizes the different knowledge bases in the different KN) IDE: build a web based agent environment, where users, software agents and groups of users can consume, communicate and share knowledge as software services 6
Email system with ZOË COMM_GENT type is also responsible for building a simple version of the shared KMO knowledge base Now only with ZOË for demonstration : Wiki Docs for Zoe software, Intertwingling your email, http://zoe.omara.ca/ Zappata personal groupware: Zoe software, Intertwingling your email, http://www.zoe.nu/ http://www.zoeprofessional.com/ later JDE Java gent DEvelopment Framework and other real agent platforms will be used together with Jena and Joseki for the KMO R&D collaboration http://jade.tilab.com/ Pilot agent implementations ZOË system provides a demonstration platform for presenting resources that are the atomic statements RDF that have been extracted from the email messages n event (and a trigger) in this KMO agent communication system is any new message that is shared and put up as atomic resources in our system the communicative actions serve goals beyond the agents internal and autonomous operation. This is due to that they are set to serve the enhancing the KMO information sharing and communication (by the external view) Conversation EMailMessage SharedMessage PersonalMessa ge News Meeting Forum Message Content Include s CalendarSched ule Include s Calendar Event 7
Conclusions the enhanced version of the ZOE based COMM_GENT will later be implemented with JDE (or some similar agent platforms and toolboxes, like KON, JCK, ZEUS, gent Factory or PTK (in PSSI project). The role of the agents is to enable and later even to automate information sharing and knowledge communities in the network Using the vertical tehnologies and toolbox if this figure we hope to connect processes for information consumption on the chosen agent platform. The suggested agent approach will provide generically an interface between the five different KMO modules KMO Modules KMO Collaboration KMO Node Portals KMO Semantic Content Management Knowledge ssimilation Organizational KMO R&D Processes Processes and Services and Services Semantic Web Toolbox Semantic Web platforms Infrastructure Server Infrastructure (web server, mail server, messaging services, web service and XML technologies) < Infrastructure > References Giovanni Caire : gent Oriented nalysis using MESSGE/UML., http://www.auml.org/auml/supplements/caire-ose2001.pdf Thomas Juan, drian Pearce, Leon Sterling: RODMP: Extending the Gaia Methodology for Complex Open Systems, http://www.cs.mu.oz.au/~tlj/rodmp.pdf David Kinny, Michael P. Georgeff: Modelling and Design of Multi-gent Systems, Lecture Notes In Computer Science; Vol. 1365 archive Proceedings of the 4th International Workshop on Intelligent gents IV, gent Theories, rchitectures, and Languages, 1-20, 1996 ISBN:3-540- 64162-9 Michael Luck, Ronald shri, and Mark d'inverno: gent-based Software Development, rtech House, 2004 Munindar P. Singh and Michael N. Huhns: Service-Oriented Computing Semantics, Processes, gents, John Wiley & Sons, Ltd., 2005 L. Uden, K. Salmenjoki, M. Hericko, L. Pavlič: Metadata for research and development collaboration between universities, Int. Journal of Metadata, Semantics and Ontologies, IJMSO, 2006 (to appear) M. Wooldridge, N. R. Jennings, D. Kinny: The Gaia methodology for agent-oriented analysis and design, Journal of utonomous gents and Multi-gent Systems 3(3): 285-312, 2000, http://www.agent.ai/doc/upload/200302/wool00.pdf 8