تامین اطلاعات برای فرآیندهای کسب و کار:اتصال جریان کار با تجزیه و تحلیل سند و بازیابی اطلاعات

Explicit modeling of business processes and their enactment in workflow systems have proved to be valuable in increasing the efficiency of work in organizations. We argue that enacted business processes — i.e. workflow management systems — form a solid basis for adequate information support in complex and knowledge-intensive business processes. To support this claim we demonstrate results from two different projects. The VirtualOffice approach employs workflow-context information to support high-precision document analysis and understanding in standard office settings; the combination of workflow context and document analysis allows for the automatic handling of incoming paper mail with respect to the appropriate workflows. The KnowMore approach focuses on the support of people who work on knowledge-intensive tasks by automatic delivery of relevant and goal-specific information. To this end, workflow context, an extended process model, and a detailed modeling of information sources are combined.

Workflow Management is a widespread technology for automating structured business processes, which can coordinate complex processes where many activities must be scheduled and dispatched among many agents. Further support comes from an integrated handling of application programs used in the process chain and a streamlined passing of application data and electronic documents flowing between different process steps. As complex business processes rely on intensive information exchange with the company's environment, they are document-driven by nature: employees deal with and react to information and knowledge transferred by and embedded in all kinds of documents, including forms, letters, books, manuals, records, either electronic or paper-based. Consequently, one would like the workflow management system (WfMS) to automatically offer access to relevant knowledge sources, or to even directly ‘pump’ information items extracted from incoming documents to the appropriate places in the data models of the actual workflow instance. This vision requires an extensive exchange of information items between and suitable semantic annotations of workflow, application, and information spaces. To realize this, the WfMS should possess interfaces for exchanging knowledge items with the surrounding support environment and be able to bridge between different conceptualizations and data models. Such considerations are not subject of today's standard WfMS approaches [12] and [3]. In order to overcome this limitation, this paper will present two different solutions, which describe how a true knowledge transfer between business processes and their surrounding information space can be established. Both approaches focus on process-embedded information delivery from documents and fit into a common description frame which is sketched in Fig. 1.The WfMS represents running business processes by workflow instances and executes them by means of a workflow engine. It is complemented by a mediator system which we call information provider. The information provider gets an information request and some additional context information from the WfMS. It accesses the documents by some kind of document index which may consist of an inverted index file as required for information retrieval tasks, but might be more sophisticated. The retrieved information is handed over to the WfMS. So, the interface between the WfMS and the information provider comprises three different kinds of information: the information request, supporting context information, and the retrieved information results supplied to the WfMS. Central to our considerations is the notion of context which allows the information provider to perform its tasks proactively and more precisely. What context information can be provided by the WfMS and which context information is required to support the information provider's job highly depends on the actual request to be fulfilled by the information provider. So, the general description frame of Fig. 1 does not help to refine the notion of context in the given scenario. Hence, we consider two concrete instantiations of the general description frame. The VirtualOffice project [18] presented in Section 2 integrates paper-based information into arbitrary workflow activities, while the KnowMore project [1], described in Section 3, supports so-called knowledge-intensive tasks (KITs) by proactive document delivery. After presenting these two projects, we review some related work and conclude with a unifying view which shows commonalities and differences of the two approaches. This unified view leads to suggestions for the design of future WfMS.

We demonstrated that the combination of business processes and their enactment in WfMSs with the information space of an organizational memory can lead to effective user support. Two different approaches have been presented to illustrate this claim. The VirtualOffice project basically aims at improving the company's information logistics by a smoother integration of paper-based documents into administrative workflows and transfer of information contained in paper documents into the electronic workflows. Thus its service can be applied virtually throughout the whole process and in each process which has interfaces to the external environment of the company where media breaks occur. It directly supports the “object level” of the business process where paper documents are an integral element. The KnowMore project is not so much focused on the object level optimization of whole and arbitrary workflows. Instead, it concentrates on knowledge-intensive core processes of a business, and KITs at the heart of those processes. Not the processes themselves are improved, but a meta level is added, an additional knowledge-service process which helps the user working on her tasks. Both examples emphasize the use of the business process models and their enactment in the WfMS as indispensable source of context information. Looking in some detail at this common goal, we can extract the following observations. Successful coupling of workflow and knowledge-intensive application requires some formal semantics as the basis of communication. That is, we need a mapping from the context information modeled in the workflow to the ontology of the information provider. The examples illustrate two principal approaches to this end: • When the workflow modeling is extended especially with respect to the support of knowledge-intensive activities, it might be possible to consider the available ontology already at process modeling time. The resulting KIT representation is then well-grounded in the ontology of the information provider. • On the other hand, if the already-defined workflow is to be left untouched, it is possible to add the necessary semantics by defining transformation rules which map the terms already used in the workflow model to the ontology. To effectively access the workflow context information it seems useful to treat the workflow instances as first order citizens in the information world. Two approaches have been presented to realize this goal: • The explicit modeling of context information in KIT variables extends the data flow model in the workflow. This is suitable if the workflow engine controls when and what is transferred outside. • The handling of a separate context pool is suitable to deal with situations where a large amount of workflow instances provide volatile context information which is required by activities outside of the workflow's influence. The continuous collection of context information and its organization according to the needs of the supporting application (e.g. the DAU component) even allows to handle context information which the WfMS only creates temporarily and which otherwise might not be available when needed. In summary, the extension of the workflow scenario is a suitable way to provide active and extended services and to transform available information into process-oriented actionable knowledge. In order to have better way for efficiently realizing intelligent services like the ones presented in this paper, without a need for clumsy work-arounds, a number of suggestions for the design of future WfMSs can be derived which are indicated in Fig. 11as an evolution of the unified view presented above: • We need a better integration of business process modeling tools and workflow engines such that all things modeled at the abstract level are also easily reflected by the enactment machinery. In the KnowMore project, for example, we had comfortable means to define sophisticated business process meta models in the ADONIS tool 8 in order to cope with KIT variables or information needs, but this is only useful for representation and simulation purposes, and cannot be compiled into operational workflows. This point would be a prerequisite for a meaningful realization of the issues following below: • It would make sense to extend business process meta models by external starting events which would be a declarative replacement of the hand-coded domain-inherent requests, creating event listener threads at system set-up time which would wait for certain events in the information space and then start the appropriate workflow. • A very important requirement would be to have more comfortable data structures (lists, structured objects, e.g. access to arbitrary object types via CORBA) for modeling workflows, i.e. for modeling workflow relevant data (used here for modeling KIT variables) as well as for specifying static call parameters for applications and information assistants. This would ease very much the communication between external assistants and the object level of the WfMS which is now realized by complex export and transformation mechanisms in VirtualOffice and by the use of a hand-made workflow engine in KnowMore.