Having access to the right information on business processes is crucial to the proper and e cient execution of all sorts of activities, such as the assessment of mortgage applications, manufacturing of goods, as well as the treatment of patients. A major challenge here is that information related to a single process is often spread out over various models, documents, and systems. This fragmentation can have disastrous consequences for an organizations operations. It can, for example, lead to delays, wastes of money, and even violations of rules and laws. The work presented in this thesis tackles these problems with algorithms that can automatically compare process information stemming from various sources. These techniques, among others, enable the detection of contradictions between the sources and improve the ability of organizations to monitor their compliance to rules and regulations.
Processes within organizations can be highly complex chains of inter-related steps,
involving numerous stakeholders and information systems [
This fragmentation occurs because organizations typically use multiple
informational artifacts to provide process information to stakeholders with di erent
informational needs or preferences [
Despite the severity of these issues, there is only limited support for organizations to e ectively deal with the negative e ects of the fragmentation of process information. Furthermore, existing techniques that provide such necessary support only focus on specific situations, involving highly-structured process information in the form of process models and event logs. Process information in less-structured formats, such as natural language documents, is largely ignored by existing work. Consequently, there is a considerable gap between the way in which organizations capture information on their processes and the support that exists for this situation. As a result of this research gap, organizations still struggle with consequences caused by the fragmentation of process information. 2
The main contributions of this thesis revolve around the definition of five techniques that focus on the comparison and alignment of process information in di erent informational artifacts. The techniques that we develop enable organizations to execute and maintain their process information more e ciently and to ensure that processes are executed according to their specification. Each technique addresses a specific scenario involving multiple informational artifacts in di erent representation formats: 2.1
Motivation. Many organizations maintain textual process descriptions alongside graphical process models. Although this makes process information accessible to various stakeholders, there is a clear risk that model and text become misaligned when changes are not applied to both descriptions consistently. For organizations with hundreds of di erent processes, the e ort required to manually identify and clear up such conflicts is considerable.
Solution. We address this problem with a technique that automatically identifies inconsistencies between a process model and a corresponding textual description. We developed tailored techniques to extract relevant process information from textual descriptions. Furthermore, our technique uses specific metrics, so-called predictors, to quantify the likelihood that the two representation formats contain inconsistencies. Evaluation. A quantitative evaluation with a collection of 53 real-world processes shows that our technique can indeed successfully identify inconsistencies in model-text pairs. For instance, we observed that our technique identifies all model-text pairs containing missing activities with a precision of 0.70. Therefore, our proposed inconsistencydetection technique can be used by organizations to quickly identify inconsistencies in their process repositories. As a result, our technique supports organizations to maintain and improve the quality of their process documentation in an e cient manner.
2.2
Motivation. Conformance checking enables organizations to automatically identify compliance violations based on the analysis of observed event data. A crucial requirement for conformance-checking techniques is that observed events can be mapped to normative process models used to specify allowed behavior. Without a mapping, it is not possible to determine if observed an event trace conforms to the specification or not. A considerable problem in this regard is that establishing a mapping between events and process model activities is an inherently uncertain task. Since the use of a particular mapping directly influences the conformance of an event trace to a specification, this uncertainty represents a major issue for conformance checking.
Solution. We developed a conformance-checking technique that avoids the problems associated with the need to select a single mapping. This is achieved by considering the entire spectrum of possible mappings generated by event-to-activity mapping techniques and capturing this spectrum in a behavioral space. Our probabilistic conformancechecking metric then provides insights into the fraction of conforming mappings, as well as useful diagnostic information.
Evaluation. A quantitative evaluation based on a collection of 557 real-world process models demonstrated that our technique can be used to obtain results in a vast number of cases where traditional conformance-checking techniques fail to do so. In particular, our technique was able to provide deterministic results for up to 50% more cases than traditional techniques.
2.3
Motivation. Textual process descriptions are widely used in organizations since they can be created and understood by virtually everyone. Because of their widespread use, they also provide a valuable source for process analysis, such as compliance checking. However, the inherent ambiguity of natural language impedes the automated analysis of textual process descriptions. While human readers can use their context knowledge to correctly understand statements with multiple possible interpretations, automated tools currently have to make assumptions about their correct meaning. As a result, conformance-checking techniques are prone to draw incorrect conclusions about the proper execution of a process.
Solution. To provide a comprehensive solution to these reasoning problems, we introduced the concept of a behavioral space as a means to deal with behavioral ambiguity in textual process descriptions. A behavioral space captures all possible interpretations of a textual process description in a systematic manner. Thus, it avoids the problem of focusing on a single, possibly incorrect interpretation.
Evaluation. We use a quantitative evaluation with a set of of 47 textual process descriptions to demonstrate that a behavioral space strikes a reasonable balance between ignoring ambiguous statements and imposing fixed interpretations on them. Furthermore, we demonstrated the usefulness of a semi-automated pruning technique to quickly reduce the level of uncertainty remaining in conformance-checking results.
2.4
Motivation. Monitoring process performance is an important means for organizations to identify opportunities to improve their operations. The definition of suitable Process Performance Indicators (PPIs) is a crucial task in this regard. Because PPIs need to be in line with strategic business objectives, the formulation of PPIs is a managerial concern. Managers typically start out to provide relevant indicators in the form of natural language PPI descriptions. Therefore, considerable time and e ort have to be invested to transform these descriptions into PPI definitions that can actually be monitored. Solution. We developed a technique that transforms an unstructured natural language PPI description into a structured notation that is aligned with the implementation underlying a business process. To achieve this, the technique builds on hidden markov models as linguistic parsers to identify the parts of a PPI description that correspond to slots in a PPI template. Then, we use semantic similarity measures and semantic constraints to fill the slots with the appropriate values belonging to particular domains. Evaluation. We evaluated the performance of our technique with a set of 129 realworld PPI descriptions from various industrial sources. The evaluation revealed that the structured PPI definitions generated by our technique are a good approximation of those created manually by experts. In particular, the technique achieves precision and recall values of, respectively, 0.89 and 0.82. Therefore, our technique represents a viable, automated alternative to an otherwise highly laborious and time-intensive, manual task. This enables organizations to more e ciently monitor the performance of their business processes and continuously adapt their monitoring activities to changing business needs.
2.5
Motivation. Process model matching provides the basis for many process analysis
techniques such as inconsistency detection and process querying. The matching task refers
to the automatic identification of correspondences between activities in two process
models. Despite the existence of a plethora of matching techniques, it has been shown
that their results leave room for improvement [
Solution. We present techniques that exploit event-log information for process model matching. instance-based process matching, which specifically focuses on information related to instances of a process. In particular, we introduce six similarity metrics that each use a di erent type of instance information stored in the event logs associated with processes. The matchers quantify similarity according to: event positions, frequencies, durations, attribute names, attribute values, and prerequisites. The proposed metrics can be used as standalone matching techniques or to complement existing process model matching techniques.
Evaluation. We demonstrated the usefulness of these similarity metrics through a quantitative evaluation using real-world data. The evaluation showed that by just considering the information specific to event logs, the introduced matchers can identify a considerable number of correspondences between event classes. This performance was particularly achieved when the six matchers were combined into a single matching ensemble. As an illustration, the results show that users would need to evaluate just 4% out of all possible correspondences in order to identify nearly 70% of the actual correspondences.
The work presented in the thesis has a variety of implications for practice and research. For practice, the work has several implications for organizations that strive to improve their organizational e ciency and process quality. It achieves this by providing insights helpful for the improved management of process information, through technical solution for the e cient resolution of inconsistencies among process representation formats, and through improved detection of business process noncompliance.
For research, the thesis laid the groundwork for the comparison and alignment of process information in semi-structured and unstructured formats. Furthermore, it provides a foundation for process analysis and reasoning in the presence of behavioral uncertainty. Finally, it presents opportunities for the improvement of process model matching techniques through the consideration of event-log information.
As such, the presented work opens up possibilities for further research that pursues the e cient use and maintenance of process information within organizations through both technological and organizational measures.