Enterprise, Business-Process and Information Systems Modeling

GPT-3 and several other language models (LMs) can effectively address various natural language processing (NLP) tasks, including machine translation and text summarization. Recently, they have also been successfully employed in the business process management (BPM) domain, e.g., for predictive process monitoring and process extraction from text. This, however, typically requires fine-tuning the employed LM, which, among others, necessitates large amounts of suitable training data. A possible solution to this problem is the use of prompt engineering, which leverages pre-trained LMs without fine-tuning them. Recognizing this, we argue that prompt engineering can help bring the capabilities of LMs to BPM research. We use this position paper to develop a research agenda for the use of prompt engineering for BPM research by identifying the associated potentials and challenges.

In order to better facilitate the need for continuous business process improvement, the application of DevOps principles has been proposed. In particular, the AB-BPM methodology applies AB testing and reinforcement learning to increase the speed and quality of improvement efforts. In this paper, we provide an industry perspective on this approach, assessing requirements, risks, opportunities, and more aspects of the AB-BPM methodology and supporting tools. Our qualitative analysis combines grounded theory with a Delphi study, including semi-structured interviews and multiple follow-up surveys with a panel of ten business process management experts. The main findings indicate a need for human control during reinforcement learning-driven experiments, the importance of aligning the methodology culturally and organizationally with the respective setting, and the necessity of an integrated process execution platform.

Modern business environments are governed by a wide range of data in various data formats. Despite the importance of integrating the data and control-flow perspective, existing business process modelling languages have only limited capability to precisely describe data-driven processes. In this paper, we propose a new approach called JSON-Nets, a variant of high-level Petri nets, that utilizes JSON technologies to integrate complex data objects in executable process models. We introduce JSON-Nets using an illustrative example and provide a formal specification, as well as a prototypical implementation of a modelling tool to evaluate our conception.

Recently, the consideration of data aspects has seen a surge in interest both from the perspective of designing processes as from a model discovery perspective. However, it seems that both research domains (models for design and model discovery) use different conceptualisations of data/object-aware systems. In an ideal situation, when (designed) models are implemented, the resulting information systems are equipped with logging functionalities that allow the rediscovery of the models based on which the information systems were implemented. However, there is a lack of guidelines on how to set up logging. From a logging perspective, logging formats are unclear about the granularity of events: the logging may be done at the level of entire tasks or at the level of the operations on individual objects, or a single log may even contain a mix of events at different granularity levels. The lack of clarity in this matter complicates the correct interpretation of log information. The goal of this paper is therefore to investigate how the concepts of object-centric logging and those for data-aware process modelling may be better aligned. This will facilitate setting up proper logging at system implementation time, and facilitate the connection of discovered models to models-for-design. The investigation resulted in iDOCEM, a meta-model that aligns the DOCEL and the Merode meta-model. Comparing iDOCEM to different other logging meta-models demonstrates that the proposed meta-model is complete enough to capture (more than) existing logging formats.

Event logs are the main source for business process mining techniques. However, not all information systems produce a standard event log. Furthermore, logs may reflect only parts of the process which may span multiple systems. We suggest using network traffic data to fill these gaps. However, traffic data is interleaved and noisy, and there is a conceptual gap between this data and event logs at the business level. This paper proposes a method for producing event logs from network traffic data. The specific challenges addressed are (a) abstracting the low-level data to business-meaningful activities, (b) overcoming the interleaving of low-level events due to concurrency of activities and processes, and (c) associating the abstracted events to cases. The method uses two trained sequence models based on Conditional random fields (CRF), applied to data reflecting interleaved activities. We use simulated traffic data generated by a predefined business process. The data is annotated for sequence learning to produce models which are used for identifying concurrently performed activities and cases to produce an event log. The event log is conformed against the process models with high fitness and precision scores.

The automatic extraction of a process model from data is one of the main focuses of a Process Mining pipeline. Decision Mining aims at discovering conditions influencing the execution of a given process instance to enhance the original extracted model. In particular, a Petri Net with data is a Petri Net enhanced with guards controlling the transitions firing in correspondence of places with two or more output arcs, called decision points. To automatically extract guards, Decision Mining algorithms fit a classifier for each decision point, indicating what path the case will follow based on event attributes. Retrieving the path followed by the case inside the model is crucial to create each decision point’s training dataset. Indeed, due to the presence of invisible activities, having multiple paths coherent with the same trace in the event log is possible. State-of-the-art method consider only the optimal path discarding the other possible ones. Consequently, training sets of related decision points will not contain information on the considered case. This work proposes a depth-first-based method that considers multiple paths possibly followed by a case inside the Petri Net to avoid information loss. We applied the proposed method to a real-life dataset showing its effectiveness and comparing it to the current state of the art.

The Internet of Things (IoT) is used in various areas to ease daily life and to make it more efficient. The IoT is a network of physical objects equipped with sensors, actuators, software, and technologies for exchanging the gathered data. In numerous domains, such as healthcare and smart home, IoT devices can be used to enable smart applications. Business Process Management (BPM), in turn, enables the analysis, modeling, implementation, execution, and monitoring of business processes. Extending BPM with IoT capabilities enhances process automation, improves process monitoring, and enables managing IoT-driven business rules. In the context of the latter the aggregation of low-level IoT data into high-level business information is a paramount important step. However, modeling, executing, and monitoring IoT-driven business rules based on BPMN 2.0 and Decision Model and Notation (DMN) might not be the best suited approach. This paper presents a framework that provides an extended BPMN notation for modeling, executing, and monitoring IoT-driven decision processes. The framework is implemented as a proof-of-concept prototype, a real-world scenario is presented to illustrate its use.

Locally distributed processes include several process participants working on tasks at different locations, e.g., craftspeople working on construction sites. Compared to classical IT environments, new challenges emerge due to the spatial context of a process. Real-time location data from Internet of Things (IoT) devices consumed through BPM technology can help businesses to implement more efficient and effective processes.

Those advantages, however, have only been touched by small parts of existing research, while the architecture and implementation of executable location-aware processes area only ever been vaguely considered. Therefore, we introduce and present a non-exhaustive list of concepts for the use of location data for BPM as well as a location-aware approach for process execution and present a multi-layer system architecture based on standard BPM technology.

Smart Contracts that embody real world legal contracts require not only a sound and secure implementation but also a careful analysis of the underlying contractual commitments. Temporal clauses are abundant in contracts, requiring permissions and obligations to be executed in temporal relationships with observed events. Before signing a contract a thorough analysis, whether breaches of temporal clauses are imminent, whether all temporal obligations can be fulfilled are inevitable to avoid the cost of violating temporal commitments. We present a contract model that focuses on modeling temporal commitments explicitly. And we present techniques based on these contract models to analyze the temporal properties of contracts, in particular, whether a party can guarantee to fulfill all temporal commitments under all foreseeable circumstances. We present a framework that supports the development and negotiation of contracts precluding the risk of violating temporal clauses.

Modelling processes with declarative process models, i.e. sets of constraints, allows for a great degree of flexibility in process execution. However, having behavior specified by means of symbolic (textual) constraints comes along with the problem that it is often hard for humans to understand which exact behavior is allowed, and which is not (think for example of checking relationships between constraints). This becomes especially problematic when modellers need to carry out changes to a model. For example, a modeller must make sure that any alteration to a model does not introduce any unwanted or non-compliant behavior. As this is often difficult for humans, editing declarative process models currently bears the risk of (accidentally) inducing unforeseen compliance breaches due to some overlooked changes in behavior. In this work, we therefore present an approach to efficiently compute the behavioral changes between a declarative process model M and a corresponding (edited) model \(M'\). This supports modellers in understanding the behavioral changes induced by an alteration to the constraints. We implement our approach and show that behavioral changes can be computed within milliseconds even for real-life data-sets.

Causality is an ubiquitous but elusive concept describing the relationship between cause and effect. In the context of business processes, it defines ontological, i.e., existential, and temporal dependencies between activities. Modeling languages define types of constraints and dependencies between activities. However, they mainly focus on the representation of the business process and the activity execution order without addressing the nature and type of activity interrelationships, i.e., its ontological profile. This paper proposes a new way of understanding activity relations through the fundamental distinction between temporal and ontological dependencies. Ten general types of activity interrelationships are derived covering all possible relationships between two activities in loop-free processes. They can be used as an aid in process redesign tasks, compliance checking, and to compare and analyze existing modeling approaches.

The paper proposes a new frontier for conceptual modeling – universal conceptual modeling (UCM) – defined as conceptual modeling that is general-purpose and accessible to anyone. For the purposes of the discussion, we envision a non-existent, hypothetical universal conceptual modeling language, which we call Datish (as in English or Spanish for data). We focus on the need for a universal conceptual data model to explain the expected benefits of UCM. Datish can facilitate the design of many different applications, including relational databases, NoSQL databases, data lakes, and artificial intelligence systems, and enable use by a broad range of users. To pave the way for rigorous development of such a language, we provide a theoretical basis for Datish in the form of a set of universal conceptual modeling principles: flexibility, accessibility, ubiquity, minimalism, primitivism, and modularity. We apply these principles to illustrate their usefulness and to identify future research opportunities.

Language technologies are gaining momentum as textual information saturates social networks and media outlets, compounded by the growing role of fake news and disinformation. In this context, approaches to represent and analyse discourses are becoming crucial. Although there is a large body of literature on text-based machine learning, it tends to focus on lexical and syntactical issues rather than semantic or pragmatic. These advances cannot tackle the complex and highly context-dependent problems of discourse evaluation that society demands. In this paper, we present IAT/ML, a modelling approach to represent and analyse discourses. IAT/ML focus on semantic and pragmatic issues, thus tackling a little researched area in language technologies. It does so by combining three analysis approaches: ontological, which focuses on what the discourse talks about, argumentation, which deals with how the text justifies what it says, and critical, which provides insights into the speakers’ beliefs and intentions, and is still being implemented. Together, these three modelling and analysis approaches make IAT/ML a comprehensive solution to represent and analyse complex discourses towards their evaluation and fact checking.

The Enterprise Architecture (EA) discipline is now established in many companies. The architectures of these companies changed over time. They resulted from a long creation and maintenance process containing processes and services provided by legacy IT systems (e.g., systems, applications) that were reasonable when they were created but might now hamper the introduction of better solutions. To handle those legacies, we started researching on the notion of EA debts, which widens the scope of technical debts to organizational aspects. However, no studies have yet been conducted to validate if the concept of EA debts has a positive influence. Within this work, we have experimented with students of an EA course. Half of the students were taught the concept of EA debts, while the other half was taught about another topic simultaneously. Afterward, the students performed a modeling task graded by EA experts among the criteria of effectiveness, comprehensibility, minimality, and completeness. The analysis revealed no significant difference between the quality of the created models by the different student groups.

The advent and growing sophistication of modern cloud-native architectures has brought into question the way we design IT infrastructures. As the architectures become more complex, modeling helps employees to better understand their environment and decision makers to better grasp the “big picture”. As the levels of abstraction multiply, modeling these infrastructures is becoming more difficult. This leads to incomplete, heterogeneous views difficult to reconcile. In this article, we present a collaborative approach focused on improving the accuracy of IT infrastructure modeling through the involvement of all stakeholders in the process. Our approach, applied in an incremental manner, is meant to increase confidence, accountability and knowledge of the infrastructure, by assigning responsibilities early in the process and leveraging the expertise of each stakeholder. It is suited for both a priori and a posteriori modeling at a low adoption cost, through adaptation of existing workflows and model reuse. Building collaborative models in such a way aims at bridging the gap between different areas of business expertise. As a result, we believe that our approach allows to perform analyses and use formal methods on larger scale models and cover wider technical domains.

This study investigates how capability maps, an Enterprise Architecture artifact, can support digital transformation and for what purposes. Despite their potential, there is currently a lack of literature on the topic. The study uses the technique of a semi-structured interview in a case study to explore the role of capability mapping in planning and executing digital transformation. The results indicate that the practice of capability mapping can play a valuable role in different phases of digital transformation. The study also provides relevant insights for organizations looking to leverage capability maps as a tool for improving their digital transformation initiatives. This is realized by describing use cases derived from the case study.

Over the last fifteen years, various frameworks for data-aware process modelling have been proposed, several of which provide a set of evaluation criteria but which differ in their focus, the terminology used, the level of detail used to describe their criteria and how these are evaluated. In addition, there are well-established evaluation frameworks of a more general nature that can be applied to data-centric process modelling too. A comprehensive and unbiased evaluation framework for (multi-)modelling approaches that also caters for more general aspects such as understandability, ease of use, model quality, etc., does not yet exist and is therefore the research gap addressed in this paper. This paper addresses this gap by using existing evaluation frameworks and developing a taxonomy that is used to categorise all the criteria from existing evaluation frameworks. The results are then discussed and related to the challenges and concerns identified by practitioners.

Design thinking is a creative process that requires brainstorming techniques that take place in a physical environment. However, such physical interactions are not possible in remote environments. In this paper, we propose a software tool for design thinking that bridges the gap between physical, digital, and virtual modeling environments. We describe and evaluate a virtual storyboarding application that enables remote collaborative design thinking in 3D and the conversion of these 3D models into 2D digital models. To evaluate the approach, we conducted an experiment with students and were able to derive directions for further research in this area.

Robotic Process Automation (RPA) offers a non-intrusive approach to workflow automation by defining and operationalizing automation rules through the graphical user interfaces of engineering and business tools. Thanks to its rapid development lifecycle, RPA has become a core enabler in many of nowadays’ digital transformation efforts. In this paper, we briefly review how some of the critical success factors of RPA endeavors can be supported by the mature techniques of model-driven software engineering (MDSE); and how RPA can be used to improve the usability of MDSE tools. By that, we intend to shed light on the mutual benefits of RPA and MDSE and encourage researchers and practitioners to explore the synergies of the two fields. To organize such prospective efforts, we define a reference framework for integrating RPA and MDSE, and provide pointers to state-of-the-art RPA frameworks.

Data visualization is an essential method for analyzing big data. Regarding the increasing demands on data visualization generation and understanding, more professional knowledge and skills are required, which are difficult to meet in practice. In most cases, people visualize data using existing templates which might not fit their requirements. We believe that it is essential to establish the connections between users’ visualization requirements and visualization solutions. In this paper, we propose a four-layer visualization framework to systematically and automatically map user requirements to data visualization solutions. Specifically, the framework is designed based on typical visual features and attributes and establishes mappings based on their semantics. Based on this framework, we have implemented a web-based prototype, which can automate the generation of visualization solutions from user visualization requirements. To evaluate the framework, we conducted a case study with one participant using the developed prototype and received positive feedback and suggestions.

In Model-Driven Engineering (MDE), source code can be automatically generated from models such as a class diagram and statecharts. However, even under the assumption that a model is correctly translated into executable code, there is no guarantee that the models correctly capture the user requirements. The validity of a model can be asserted by means of model execution or testing the (prototype) application generated from the model. The completeness of such validation effort can be expressed in terms of model coverage of the executed scenarios. TesCaV is a Model-Based Testing (MBT) tool that provides users feedback by visualizing which test cases have been performed and which ones not yet. This allows TesCaV to be used in an educational setting as its feedback about the manual test cases can be alleviated to let students understand how to adequately test a software system. However, it remains unclear what the best way is to provide this feedback in terms of providing the user maximal information with minimal cognitive load. This research evaluates several proposed visualizations created according to information visualization principles, and makes a ranking based on a questionnaire distributed to 45 participants.

Process discovery aims to discover models to explain the behaviors of information systems. The Inductive Miner (IM) discovery algorithm is able to discover process models with desirable properties: free-choiceness and soundness. Moreover, a family of variations makes IM practical for real-life applications. Due to the advantages, IM is regarded as the state of the art and has been implemented in commercial process mining software. However, IM can only discover block-structured process models that tend to have high fitness but low precision. To improve the quality of process models discovered by IM while preserving desirable properties, we propose an approach that applies property-preserving (free-choiceness and soundness) reduction/synthesis rules to iteratively modify the process model. The experimental results show that the models discovered by our approach have a more flexible representation while preserving desirable properties. Moreover, the model quality, as measured by the F1-score, is improved compared to the original models.