Research On Key Technology Of Ontology Based Workflow Centric Collaboration

As widly use of Web technology, now a days, the collaboration systems are implemented based on Web technology more and more. They utilize the open characteristic of Internet to share resource widely. In that kind of open computing environment, the computing resources that systems depend on are diversiform and ever changing, that request the systems must be flexible enough to adjust the key elements to adapt to the change of environment. Meanwhile, as the business processes are used in open environment, the adaptabliliy is required by process essentially. The business processes should be able to be transformed dynamically according to the changing of requestment of business, to be suitable for the requirement of diversity of business. The diversity is hard to be defined by enumerating all the situations, or the modeling of business process would be too complex, and that would add the difficulty to the business management system.This dissertation talks about the key techniques of ontology based workflow centric collaboration system, such as ontology modeling, semantic query, and business object management. The aim is to make foundations for construction of a scalable, user customizable and intelligent business characterized collaboration system. We use ontology theory and techniques to model key elements of collaboration activities, by describing concepts and relationship between them accurately and semanticly to represent the collaboration activities related knowledge like objects, resource, and business rules.In order to reduce the complexity of moding and realize the dividing of modeling work, the ontology repository is constructed in three levels, which are Static Abstract Level, Static Concrete Level, and Dynamic Concrete Level.The Static Abstract Level ontologies are mainly those that describe basic common attributes and rules of the collaboration system, as well the essential relations between them. In fact, ontologies at this level define the system frame-work; The Static Concrete Level ontologies are more concrete than those at Static Ab-stract Level. They are usually subclasses or instances of those abstract ontologies. They depict the domain related classes and models. The procedures that we build up ontologies at this level are exactly the procedures that we make the system customized to special domain application; the ontologies at Dynamic Concrete Level are mainly those runtime instances that the upper levels ontologies required to be included into the repository. In fact they together constitute the system runtime context. Our system is based on ontology to define system - framework, which consists of four sub-models: organization ontologies, resource ontologies, process ontologies, and context ontologies. By modeling according to different domain relative degree abstraction layers, we can add the collaboration knowledge into the system dynamically. The models are easy to be modified and extended and that enable the system to be customized.Process models are main gists that colleberation engine flows. they define the behavior of the system. The main task of engine is to realize those definitions. The instances, also called tasks, are created by users or triggered by some kind of system events. The engine needs to access the definitions of the corresponding models of the instances, drives the process to flow, and performs the operation difined by the process models.We must utilize the knowledge added into the system by definition to guide and constrain the collaboration activities. As for the flow driven mechanism, the workflow engine controls the behavior of the system according to the definition of business process models. The engine interacts with ontology relationship discovering reasoner to access the collaboration knowledge of the system.Reasoning ability based the ontology models is the primary characristic of the ontology based collaboration system distinguishing with other systems. The framework rules of the collaboration system and domain independent knowledge are stored in the system in ontologies forms. To utilize that knowledge, we must use the reasoning function of the system. The system modelizes the ontoloty by aimimg at the characteristic of OWL DL, and stores the ontology models into the database in the forms of RDF graphs, and performs semantic quiring by using Description Logic reasoning. When the engine is processing the business flow, it would need to confirm some relationship and object dynamically, or make some judgement by deducing. Then the engine must call the semantic searching function of reasoner. The reasoner reasons to discover implicated relationship of runtime context to answer the calling of engine.The flow engine is the core module of the workflow centric collaboration system. It is in charge of the management of the system running. The engine drives the instances in the system to startup, flow, hang up, and end; distribute tasks to users according to definitions of organization models, privilege models, resource models, and process models, represents the business data to the users, and performs transaction on business data under the defined constrain rules according to the definition of process models and recource models, as well as the runtime context.The engine works according as Petri net scheduling algorithm. The kernel scheduling mechanism can be described as flow: whenever state Token of a Node transported or changed, re-browse all the Nodes of the flow instances. There are two types of Nodes, State Nodes and Activity Nodes. State Nodes represent for Places of Petri net, and Activity Nodes represent for transitions of Petri net.The business objects those tasks handled with are domain dependent. They are focuses of domain collaboration. In traditional workflow system, in order to enable the workflow engine with generality, the engine had to be domain independent. Thus the process definition is deviated with the business data. All the operation on the business data are performed at client. Because of that, the engine lacks with the ability of data management. But at most cases, the performing of task would bring effect on business data. If the engine can not handle this kind of changes, we must communicate with interface between engine and client frequently to exchange the changes of states. That puts addition complexity on client and reduces the scalability of the system.By modeling the business object with ontology, describing the relations between business objects and relations between business objects and tasks, we can enable the engine the ability of performing operations that process models specified on the business objects according to the runtime context. The operations performed on the business data are mainly adding, deleteing, computing, and modifying.Dynamic workflow management is also a target of this dissertation. It enables the system to get possible routes and tasks according to the runtime context. After that, the system can compute with parameters that user chosen and routing rules to decide the route direction flexibly and ensure process models can be chozen and modified at runtime when exception occoured during workflow instances running. To resolve the problems of dynamic, flexibility, and scalability of workflow, we use ontoloty to -describe domain related information, and adapt the idea of components replacement, use Worklet to provide system with choosable sub process definitions, use ripple down rules to support dynamic rules choices.Worklet is a scalable self-comprised sub process. It can be used in many scenes depending on special tasks context. In short, a worklet is a relative integrated process definition. It is deferent with normal process definitions because it is used in a larger process definition to handle some special problem, and its emergence depends on special context. RDR is a gradual knowledge acquiring technology; also it is a rule representation mode. It is a user centric knowledge depository system construction mothod. It allows users or domain experts to buid up the knowledge base quickly. RDR method simplifieds knowledge retrieving as follow: it distributes a Conclusion to a Case by confirm difference between current Case and Cornerstone Case. In that way, users can establish their exception handling rule tree, and use exception handling meta-language to set up exception handling process.Based on the works that stated above, we prepare technic for a collaboration system that has semantic characteristic and is easy to be extended dynamically.