Fragment-based Case Management (fCM) is an approach to handle flexible, knowledge-intensive business processes. An fCM model is composed of four artifacts, modeling process behavior and data. Diferent modeling languages and hidden dependencies among these artifacts make modeling especially challenging. However, no adequate tooling that supports designers exists. To close this gap, we propose fcm-js, a modeling tool that provides a joint, visual user interface for all artifacts and integrates automated guideline checking based on fCM guidelines.
Recently, knowledge-intensive processes receive more and more attention in the BPM
community. Knowledge-work is highly flexible and driven by the decisions of experts (e.g., physicians).
Existing BPM approaches provide insuficient support for these processes [
While tools for modeling with process-centered languages, such as BPMN [
In this demo, we present the open-source fCM modeling tool fcm-js1. It enhances the modeling process by ofering an intuitive user interface and improves the model quality by integrating several fCM modeling guidelines. fcm-js allows the user to model all artifacts of an fCM model at the same time, which gives a good overview of the progress and can speed up modeling. Furthermore, fcm-js includes an extensive guideline checking system. Currently, 20 guidelines regarding good fCM modeling2 are implemented on diferent levels. Users are informed of 1 2 potential errors and best practice violations in their artifacts, and they are ofered quick-fixes. Moreover, some guidelines regarding the consistency between the artifacts are enforced by the tool, thereby ensuring a certain degree of quality. In the remainder, we present the main functionalities in more detail and discuss the maturity of the tool. For this, a small user study is presented, and future work is highlighted. 2. Overview and Features fcm-js is a web application mainly written in HTML and JavaScript. One of its core features is that it is composed of one modeler for each artifact of fCM models. Figure 1 shows a screenshot with all important visual components: the modelers for (1) fragments, (2) data model, (3) goal state, and (4) object lifecycles (OLCs); (5) the guideline violation table. In the following, the modelers are described in more detail, and we briefly explain the implementation of two other core features, namely consistency and guideline checking.
The individual modelers are technologically based on diagram-js, a “toolbox for displaying and modifying diagrams on the web”3. Hence, they are designed similarly: as user interface, SVG-based diagram canvases are provided, featuring drag-and-drop-based visual modeling and advanced features such as copy and paste. As interface for custom internal modules and auxiliary components, each modeler features an event bus, which can be used to listen to, handle, and fire various events. It is the key component to communicate with the modelers on a software level.
The Fragment Modeler builds upon the BPMN modeler bpmn-js4, as fCM fragment models use a subset of BPMN. The Fragment Modeler supports users, for example, when placing data object references by making them choose data classes and states from the data and object lifecycle models. The Data Modeler allows creating basic UML class diagrams consisting of classes with names and attributes, and associations with multiplicities and fCM-specific goal multiplicities. In the Goal State Modeler, states from the OLCs can be put together in disjunctive normal form with dropdown menus. The Object Lifecycle Modeler allows selecting a data class whose OLC should be displayed. OLCs are modeled exclusively with states and unlabeled transitions.
One inherent challenge of fCM modeling is having to work in four models at the same time, and ensuring their consistency with each other.
fcm-js displays all four modelers side by side (cp. Fig. 1). It allows users to adjust the spacing on-the-fly and select which modeler to display “in focus” on the left side. This way, users can seamlessly work on multiple artifacts at the same time, avoiding mental context switches.
Furthermore, fcm-js enforces general consistency between the artifacts. To this end, we put a mediator component in place, which connects to each event bus. For instance, deleting and renaming data classes or OLC states is captured and propagated across the modelers, which can then adapt their models automatically.
All artifacts can be imported and exported as one file, allowing users to save their models and continue modeling later on.
Modeling guidelines help designers to detect errors and to assess the quality of a model. However, checking guidelines manually is tedious and error-prone. With the mediator component, fcm-js already automatically enforces most consistency-based guidelines from the fCM guideline catalog5 . For many other guidelines, fcm-js features automated checking. Guidelines are translated into one or more checks, which are unambiguously met or violated by a given model. These checks are reevaluated whenever the user changes the model.
As an example, consider the guideline with catalog ID C36, which states that state transitions induced by activities should be covered by the OLCs. Furthermore, consider the process fragment displayed in Fig. 2, which shows the activity Invite Speakers that changes instances of class Conference from state reviewing closed to state speakers invited. If there is no matching transition in the OLC of class Conference, this fragment violates C3. In fcm-js, this violation is detected, and the user is made aware of it by highlighting the activity (cp. Fig. 2) and listing the violation 4https://github.com/bpmn-io/bpmn-js 5https://github.com/bptlab/fCM-design-support/wiki 6https://github.com/bptlab/fCM-design-support/wiki/Consistency#c3 in a hideable guideline violation table (cp. (5) in Fig. 1). For each violation, fcm-js provides an explanatory message, and optionally a list of quick-fixes, i.e., concrete actions to fix the violation that can be applied automatically. In the example, clicking on the first option would automatically create the missing OLC transition for the user, which solves the violation.
Immediate feedback through automated violation detection and highlighting helps designers to learn fCM in a trial-and-error fashion. Furthermore, quick-fixes accelerate the user workflow by resolving violations more quickly.
While fcm-js is still active in development, it is already in a stable version and can be used in projects. In this section, we will first showcase two small user studies conducted with the tool and then provide an outlook for future work.
We assessed the efectiveness of fcm-js though a user study comparing the quality of fCM models created by hand with models created with fcm-js. Two groups of each five competent fCM designers modeled the same knowledge-intensive process. One group received the guideline catalog and fcm-js, the other did not. Manually checking created models showed an increase of fulfilled guidelines in the tool-supported group (66% vs. 84%). This efect was even higher when only considering guidelines that were implemented in fcm-js (66% vs. 91%). With that, we conclude that fcm-js helps designers to fulfill guidelines and thus produces fCM models of higher quality.
Furthermore, we conducted think-aloud sessions and interviews with five diferent competent fCM designers, who were asked to model an fCM case model based on a textual description using fcm-js. The participants stated that they felt more confident in modeling and rated their models as higher in quality compared to previous modeling procedures. In particular, the general tool structure and automatically performed consistency checks were positively highlighted. We infer that fcm-js improves the fCM modeling experience. Our tool is a stand-alone web application that focuses on the design-time support and aims to make fCM more accessible to novice users. Future work could consider an integration into existing fCM execution pipelines. Regarding the expandability of fcm-js, more guidelines can be easily added through the unified guideline interface.
Furthermore, user experience can be improved. For instance, undoing and redoing are currently not possible in fcm-js and could be a significant enhancement. However, especially considering our strict consistency guidelines which automatically synchronize all artifacts, implementing un- and redoing is a challenging task.
In general, the design of fcm-js can be transferred to future projects on hybrid modeling approaches. We want to stress especially the side-by-side layout and the consistency and guideline checking, which helped designers to cope with the complexity induced by composed models.