=Paper=
{{Paper
|id=Vol-2350/paper20
|storemode=property
|title=Towards an Assistive and Pattern Learning-driven Process Modeling Approach
|pdfUrl=https://ceur-ws.org/Vol-2350/paper20.pdf
|volume=Vol-2350
|authors=Emanuele Laurenzi,Knut Hinkelmann,Stephan Jüngling,Devid Montecchiari,Charuta Pande,Andreas Martin
|dblpUrl=https://dblp.org/rec/conf/aaaiss/LaurenziHJMPM19
}}
==Towards an Assistive and Pattern Learning-driven Process Modeling Approach ==
https://ceur-ws.org/Vol-2350/paper20.pdf
Towards An Assistive and Pattern Learning-driven
Process Modeling Approach
Emanuele Laurenzi, Knut Hinkelmann, Stephan Jüngling,
Devid Montecchiari, Charuta Pande and Andreas Martin
FHNW University of Applied Sciences and Arts Northwestern Switzerland
School of Business, Riggenbachstrasse 16, CH-4600 Olten, Switzerland
{emanuele.laurenzi|knut.hinkelmann|stephan.juengling|devid.montecchiari|charuta.pande|andreas.martin}@fhnw.ch
Abstract since they lack of modeling know-how. They neither
The practice of business process modeling not only requires understand the many BPMN syntax elements nor know how
modeling expertise but also significant domain expertise. to design correct process models with a good style.
Bringing the latter into an early stage of modeling contributes
to design models that appropriately capture an underlying
We tackle this problem by proposing an assistive
reality. For this, modeling experts and domain experts need
to intensively cooperate, especially when the former are not modeling approach that makes use of validated domain-
experienced within the domain they are modeling. This specific semantic process patterns, that are established over
results in a time-consuming and demanding engineering time to compensate the lack of business process modeling
effort. To address this challenge we propose a process expertise of domain experts.
modeling approach that assists domain experts in the creation
The idea of using design patterns as reusable solutions for
and adaptation of process models. To get an appropriate
assistance, the approach is driven by semantic patterns and common design problems could successfully be transferred
learning. Semantic patterns are domain-specific and consist from the domain of physical construction of cities
of process model fragments (or end-to-end process models), (Alexander et al., 1977) to the domain of object oriented
which are continuously learned from feedback from domain software design (Gamma, 1995). We reuse this idea for the
as well as process modeling experts. This enables to
construction of a semantic repository containing both
incorporate good practices of process modeling into the
semantic patterns. To this end, both machine-learning and domain-specific patterns and business process models in the
knowledge engineering techniques are employed, which form of an ontology.
allow the semantic patterns to adapt over time and thus to Our approach goes beyond the mere syntactical or
keep up with the evolution of process modeling in the semantic validation, which is already implemented in
different business domains.
current business process modelling tools. With the support
of machine learning techniques we aim to learn patterns of
Introduction good business process modeling that typically reside in
modeling and domain expert’s heads.
The importance of having good process models, which
accurately describe the implemented or intended business
processes of a company, is well recognized. Nevertheless, Learning Domain-Specific Semantic Patterns
many companies face similar practical problems during the
In software engineering, the adoption of design patterns has
analysis and design phase of their business process
shown particular success in assisting programmers to
management. The domain experts, which are the owners and
develop software.
main users of the business process repository, mostly
In enterprise modeling (which includes the practice of
delegate the analysis, design and construction of business
BPM), the use of patterns originates from the field of graph
process models to business analysts or to the IT
theory addressing the problem of graph pattern matching
departments, who have the necessary modeling and
(Fu, 1995).
engineering skills. This leads to the additional effort for
The research in establishing and using patterns to support
communication and collaboration, because modeling
the process modeling is still quite active (Deng et al., 2017;
experts may lack the necessary expertise from the specific
Delfmann et al., 2010; Gruhn & Laue, 2009). However,
business domain. The domain experts delegate the modeling
these patterns are quite generic and on an abstract level, so
Copyright held by the author(s). In A. Martin, K. Hinkelmann, A. Gerber, Engineering (AAAI-MAKE 2019). Stanford University, Palo Alto,
D. Lenat, F. van Harmelen, P. Clark (Eds.), Proceedings of the AAAI 2019 California, USA, March 25-27, 2019.
Spring Symposium on Combining Machine Learning with Knowledge
that the reuse and application is mainly promoted within the build a sufficient and appropriate set of domain-specific
modeling expert community which have the ability to semantic patterns. Furthermore, the process reality to be
identify the patterns and to apply them in the target domain. modeled lays in domain experts’ heads but domain experts
In most cases, the domain experts do not have the expertise are not necessary aware of. This means that the knowledge
to adapt the generic patterns to their business domain. is tacit (Polanyi, 1966). Learning how to identify a domain-
Hence, ideal patterns should aim to stay in the background. specific semantic pattern promises to support this
This means, a pattern should not only be assistive in terms knowledge extraction process (see lower arrow in Figure 1).
of (a) syntactic and (b) semantic correctness but also help The need to learn domain-specific semantic patterns led
with appropriate: to formulate the following research question:
c) level of abstraction, which fits the purpose of the
- How to learn domain-specific semantic patterns to
process model granularity
support the good process model practice?
d) content (i.e., terminology) fitting to the domain
targeted by the process model and To address this question, we start introducing the three
e) modeling styles and conventions, e.g., see learning sources that are typically adopted as best practices
scenarios in (Silver, 2011). to create a good process model.
In this paper, we define this kind of pattern as a domain-
specific semantic pattern for good process modeling. Learning Source Cases
Several research work evolved around the term semantic
patterns (Staab et al., 2001; Saif et al., 2014; Soffer et al., We distinguish between three sources of cases from which
2007) (also known as ontology design patterns (Damjanovic we can learn domain-specific semantic patterns for good
& Violeta, 2009)). However, different from us, they adopt a business process modeling:
knowledge engineering approach, which focuses on explicit (a) Modeling expert that give feedback to domain
knowledge that is available from the process models (i.e., experts’ models
syntax and semantic aspects). In our approach, we (b) Simulation of models
incorporate a learning approach to deal with tacit knowledge (c) Experience with process executions.
(see left hand side of Figure 1).
The first learning source typically refers to
(a) syntax
(b) modeling style
(c) level of details of a business process, i.e.,
abstraction level, and
(d) business process description.
Modeling style and conventions come from experience of
experts and are documented, for example, in (Silver, 2011)
and Schallert & Rosemann (2012). Syntactical constraints
are also documented in the BPMN specifications (OMG,
2011) and implemented in some modeling tools. For
example, commercial BPMN tools such as Camunda,
Signavio, Bizagi, or Flowable implement syntactical checks
We argue that the appropriate identification and that validate syntactical correctness of the design of BPMN
construction of domain-specific semantic patterns can be models. Both syntax and modeling style can be
resolved by combining knowledge engineering approaches implemented through the knowledge engineering task.
with learning approaches. To elaborate on this we refer to However, learning them would relieve the engineering
the different types of knowledge known from knowledge effort especially for the modeling style and conventions,
management that are shown Figure 1. which sometimes are subjective or application domain-
The identification of an appropriate domain-specific dependent.
semantic pattern aims to capture the underlying business Feedback regarding the abstraction level of a business
reality as an appropriate process model. process (c) relies on the intuition and experience of the
Based on their experiences, modeling experts—maybe process modeler, i.e., tacit and self-aware knowledge,
together with domain expert—are able to construct domain- respectively. Thus, it can be dealt with a learning approach.
specific semantic patterns. This is regarded as a knowledge Similarly, feedback regarding the business process
engineering task (see upper arrow in Figure 1). However, it description (d) relies on experience of the process modeler.
would result in a tremendous engineering effort to manually Sometimes, the usable terms within a specific project or
application domain are documented, which makes it an appropriate engineering of a similarity configuration is a
alternative source to the feedback. Whereas the former critical requirement for applying CBR.
knowledge as implicit but of self-aware type, the latter, Martin (2016) introduced an approach, called ICEBERG,
which is explicit and belongs to the documented knowledge how ontology-based case-based reasoning (OBCBR) can be
type. applied in process execution by comparing cases of process
The second learning source (model simulation) refers to fragments. Martin (2016) pointed out that the engineering of
the behavior of process models, also known as behavioral the similarity configuration is a critical step and allocates
semantics (Muzi et al., 2018 ; Corradini et al., 2017; significant resources from domain experts. Martin (2016)
Mendling, 2009). The most widespread approach in process and Martin & Hinkelmann (2018) introduced a procedure
modeling consists of mapping a process model to a formal model for the design and implementation of an OBCBR.
semantic like Petri Nets (van Dongen et al., 2008). The Certain procedural stages are essential in this context of
approach is used to identify the existence of deadlocks or configuring a process pattern similarity model as well:
live-locks through simulation of the correspondent Petri Net
1. At a first stage, it is essential that domain experts
model. For example, in the tool BProve (Corradini et al.,
decide on an enterprise-specific conceptualization
2017), Petri Nets are used to simulate the execution of a
or nomenclature to build a process fragment
business process model to assess the safeness and soundness
characterization vocabulary or feature set.
of BPMN collaboration, to check both the existence of
2. Then the various mental similarity and adaptation
deadlocks and proper completion of BPMN models.
models need to be elicited and externalized from
Issues like deadlocks or live-locks can be learned by
the domain experts.
running one of the existing dedicated tools as they
3. Next definition of the process fragment
sometimes remain difficult to detect even for modeling
characterization will be implemented within an
experts.
enterprise, domain and/or modeling ontology.
The third learning source, (3), refers to the improvement
4. Finally, knowledge and domain experts configure
of a process models based on the analysis of the run-time
the similarity model as introduced by Martin et al.
later execution of the business process that can occur
(2017) within the ontology. This configuration is
manually by the modeling experts or through the support of
made by determining global and local similarity
process mining tools (van der Aalst, 2009).
functions and assigning weights.
This list of learning sources reveals that there is still a
quite significant amount of implicit knowledge laying in By describing this approach we applied knowledge-based
expert’s heads, which is neither documented nor and knowledge engineering methods in combination and
implemented in process model tools. By tackling the simultaneously. However, there are some drawbacks. The
challenge of explicating such knowledge through learning first one concerns the high effort for engineering the
approaches, we want to show how we intend to address the characterization, which can be defused by establishing a
above introduced research question. semantic repository. Secondly, and more difficult to tackle,
Thus, in the following three sections we introduce three humans are not good in selecting appropriate similarity
ways of learning a domain-specific semantic pattern: (1) features and have difficulties in estimating similarity
Learning Process Fragment Similarity Model, (2) Learning weights.
Pattern’s Abstraction Level (3) Learning Pattern’s A possible way to overcome the mentioned difficulties in
Description. Each of them presents one machine learning selecting similarity features and weights could be taken by
approach, which builds on existing techniques. prior execution of a data-driven machine learning approach.
In one of our previous works (von Rohr et al., 2018) we
could show that an initial data-driven machine learning
Learning Process Fragment Similarity Model approach based on a regression model can be used to
Likewise, in case-based reasoning (CBR) a similarity model generate, respectively derive similarity features and the
is an essential component. CBR is known as a technically corresponding weights.
independent methodology (Watson, 1999) for humans and This section shows how a similarity model can be
information systems to reason by remembering (Leake, engineered, how it can be derived and embedded into a
1996). During remembering previous cases will be knowledge-based environment, and finally how such a
compared by applying a similarity model. The ultimate goal model can be learned by applying a data-driven machine
in CBR is then, to transfer or adapt the knowledge from learning approach. The resulting similarity model allows
previous cases to the current situation/problem. CBR has its retrieving the most similar process model to the one being
roots in cognitive science, machine learning and knowledge- designed. Later, feedback of domain experts could then re-
based systems (Martin & Hinkelmann, 2018). The initialize the data-driven machine learning procedure to
determine the adaptation of the similarity measure over Learning Pattern’s Description
time.
In a related work, Wasser & Lincoln (2012) proposed a
Process Descriptor Catalog (PDC) to describe activities
Learning Pattern’s Abstraction Level within a business process according to two main elements,
namely Object and Action, and four taxonomies, namely an
By using the similarity model, it is possible to learn the
Action Hierarchy Model (AHM), an Object Hierarchy
abstraction of a pattern found in several similar process
Model (OHM), an Action Sequence Model (ASM) and an
fragments. As an example, let us consider the process
Object Lifecycle Model (OLM). These models were used to
fragment in Figure 2. It models the chain of actions involved
organize a set of virtual activities within a particular process
in sending an invoice.
domain to express the relationships between actions and
object both hierarchically and in term of execution order.
Check Send Check
Given as input a business process, a Natural Language
invoice invoice payment
Processing (NLP) system was used to derive the actions and
Figure 2: Pattern to send Invoice objects involved in the process. By mapping this
information to the models, it was possible to assess the
This is a semantic sequence that can be observed in
similarity between the given process and a possible set of
almost all the processes that involve sending invoices. Such
semantically correlated new activities. Then, the choice of
a pattern can be considered as an abstraction that can be
the modeler was used as feedback to learn to provide better
extended in a similar process model being designed as
suggestion adapting the similarity function called in the
shown in Figure 3.
study, distance function.
Extending the approach of activity descriptors to the
Check Send Check description of fragments of processes would fit well with the
monthly monthly payment
invoice invoice intention of learning pattern’s description.
Figure 3: Extension of the Send Invoice pattern
Future Work and Conclusion
Check Send Check
document document payment
In our research roadmap, we intend to implement the three
presented machine learning approaches in the form of
Figure 4: Abstraction of the Send Invoice pattern functionalities of our Agile and Ontology-Aided Modeling
Environment (AOAME) (Laurenzi at al., 2018) (see Figure
It is also possible that the similarity model learns different 5). The latter builds on the ontology-based meta-modeling
levels of abstraction for the same pattern over a period of concept described in (Hinkelmann et al., 2018) and
time, as illustrated in Figure 4. Whether the learned seamlessly integrates models with ontologies. Therefore,
abstraction of the pattern is useful or not can be verified techniques for semantic annotations/lifting or
through the feedback from the domain experts. The model transformations are not needed as the semantic repository
in Figure 4 shows an example of an abstraction that cannot (i.e., see right-hand side of Figure 5) is built while modeling
be used, as sending every type of document does not lead to takes place.
a payment. Thus, there is a need for a mechanism to The semantic repository contains ontologies reflecting
incorporate the positive or negative feedback into a pattern both business process models (i.e., Model Ontology)
repository. CBR has limitations in this respect, as it will designed by the domain experts and the learned domain-
create a new pattern based on the feedback but not adapt the specific semantic patterns, i.e., Pattern Ontology. The latter
level of abstraction of the same pattern. We explore how contains either an entire process model or fragments of
CBR can be augmented with Reinforcement Learning (Pack process models with specific modeling style, conventions,
Kaelbling et al., 1996) as a mechanism to incorporate the behavioral aspects, abstraction level, etc. It is out of scope
domain experts' feedback – a useful level of pattern of this paper to precisely define what such a pattern consists
abstraction will achieve a positive reward, and an incorrect of.
abstraction will achieve a negative reward or a penalty. The learning for the good business process modeling
Based on the rewards or penalties, the similarity model will comes from the feedback of a variety of different sources.
be able to identify the right level of abstraction for a process First, the system can learn and adapt based on the rating of
model being designed.
Figure 5. The Assistive and Pattern Learning-driven Approach. Adapted from (Laurenzi et al., 2018)
the modeling experts that give feedback to the models being Deng, S., Wang, D., Li, Y., Cao, B., Yin, J., Wu, Z., & Zhou, M.
designed by the domain experts. Second, the process models (2017). A Recommendation System to Facilitate Business Process
Modeling. IEEE Transactions on Cybernetics, 47(6), 1380–1394.
can be simulated during design time, and the result of the https://doi.org/10.1109/TCYB.2016.2545688
simulation can provide fast feedback to the domain experts
Emmenegger, S., Hinkelmann, K., Laurenzi, E., Martin, A.,
while modeling. Third, the domain experts can manually Thönssen, B., Witschel, H. F., & Zhang, C. (2017). An Ontology-
give feedback based on their experience of later process Based and Case-Based Reasoning Supported Workplace Learning
implementation, and the modeling experts can give Approach. In Communications in Computer and Information
feedback based on the results from process mining. All of Science (pp. 333–354). Springer, Cham.
https://doi.org/10.1007/978-3-319-66302-9_17
these sources are used to learn best practices and
continuously improve the Pattern Ontology. Emmenegger, S., Hinkelmann, K., Laurenzi, E., & Thönssen, B.
(2013). Towards a Procedure for Assessing Supply Chain Risks
The semantic repository has the benefit of having process Using Semantic Technologies. Communications in Computer and
models and patterns defined in a machine interpretable Information Science (Vol. 415). https://doi.org/10.1007/978-3-
representation. Thus, reasoning services and semantic rules 642-54105-6_26
can be applied to deduce new knowledge. This was already Fu, J. (1995). Pattern matching in directed graphs. In Galil Z. &
proven to be successful in several research works in Ukkonen E. (Eds.), Combinatorial Pattern Matching (pp. 64–77).
different domains, e.g., Business-IT alignment in the Cloud Springer, Berlin, Heidelberg. https://doi.org/10.1007/3-540-
60044-2_35
(Kritikos et al., 2018; Hinkelmann et al., 2016), workplace
learning (Emmenegger et al., 2017), and supply chain risk Gamma, E. (1995). Design patterns : elements of reusable object-
oriented software. Addison-Wesley.
management (Emmenegger et al., 2013).
Gruhn, V., & Laue, R. (2009). Reducing the cognitive complexity
of business process models. In 2009 8th IEEE International
Conference on Cognitive Informatics (pp. 339–345). IEEE.
References https://doi.org/10.1109/COGINF.2009.5250717
Alexander, C., Ishikawa, S., & Silverstein, M. (1977). A pattern Hinkelmann, K., Laurenzi, E., Lammel, B., Kurjakovic, S., &
language : towns, buildings, construction. Oxford University Woitsch, R. (2016). A Semantically-Enhanced Modelling
Press. Environment for Business Process as a Service. In 2016 4th
Corradini, F., Fornari, F., Polini, A., Re, B., Tiezzi, F., & Vandin, International Conference on Enterprise Systems (ES) (pp. 143–
A. (2017). BProVe: A formal verification framework for business 152). IEEE. https://doi.org/10.1109/ES.2016.25
process models. In 2017 32nd IEEE/ACM International Hinkelmann, K., Laurenzi, E., Martin, A., & Thönssen, B. (2018).
Conference on Automated Software Engineering (ASE) (pp. 217– Ontology-Based Metamodeling. In Dornberger R. (Ed.), Business
228). IEEE. https://doi.org/10.1109/ASE.2017.8115635 Information Systems and Technology 4.0. Studies in Systems,
Damjanovic, V., & Violeta. (2009). Ontology design patterns for Decision and Control. (pp. 177–194). Springer, Cham.
the semantic business processes. In Proceedings of the 4th https://doi.org/10.1007/978-3-319-74322-6_12
International Workshop on Semantic Business Process Kritikos, K., Laurenzi, E., & Hinkelmann, K. (2018). Towards
Management - SBPM ’09 (pp. 51–54). New York, New York, Business-to-IT Alignment in the Cloud. In Z. Mann & V. Stolz
USA: ACM Press. https://doi.org/10.1145/1944968.1944977 (Eds.), Advances in Service-Oriented and Cloud Computing.
Delfmann, P., Herwig, S., Lis, Ł., Stein, A., Tent, K., & Becker, J. ESOCC 2017. Communications in Computer and Information
(2010). Pattern Specification and Matching in Conceptual Models Science (pp. 35–52). Springer, Cham. https://doi.org/10.1007/978-
- A Generic Approach Based on Set Operations. Enterprise 3-319-79090-9_3
Modelling and Information Systems Architectures (EMISAJ), 5(3), Laurenzi, E., Hinkelmann, K., & van der Merwe, A. (2018). An
24–43. https://doi.org/10.18417/emisa.5.3.2 Agile and Ontology-Aided Modeling Environment. In R.
Buchmann, D. Karagiannis, & M. Kirikova (Eds.), The Practice of Systems Engineering (pp. 450–464). Springer-Verlag.
Enterprise Modeling. PoEM 2018. (pp. 221–237). Vienna: https://doi.org/10.1007/978-3-540-69534-9_34
Springer, Cham. https://doi.org/10.1007/978-3-030-02302-7_14 von Rohr, C. R., Witschel, H. F., & Martin, A. (2018). Training
Leake, D. B. (1996). CBR in Context: The Present and Future. In and Re-using Human Experience: A Recommender for More
D. B. Leake (Ed.), Case-Based Reasoning: Experiences, Lessons, Accurate Cost Estimates in Project Planning. In Proceedings of the
and Future Directions (pp. 1–35). Menlo Park: AAAI Press/MIT 10th International Joint Conference on Knowledge Discovery,
Press. Knowledge Engineering and Knowledge Management (pp. 52–62).
Martin, A. (2016). A combined Case-based Reasoning and Process SCITEPRESS - Science and Technology Publications.
Execution Approach for Knowledge-Intensive Work. University of https://doi.org/10.5220/0006893200520062
South Africa. Wasser, A., & Lincoln, M. (2012). Semantic Machine Learning for
Martin, A., Emmenegger, S., Hinkelmann, K., & Thönssen, B. Business Process Content Generation. In Meersman R. et al. (Ed.),
(2017). A viewpoint-based case-based reasoning approach On the Move to Meaningful Internet Systems: OTM 2012 (pp. 74–
utilising an enterprise architecture ontology for experience 91). Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-
management. Enterprise Information Systems, 11(4), 551–575. 642-33606-5_6
https://doi.org/10.1080/17517575.2016.1161239 Watson, I. (1999). Case-based reasoning is a methodology not a
Martin, A., & Hinkelmann, K. (2018). Case-Based Reasoning for technology. Knowledge-Based Systems, 12(5–6), 303–308.
Process Experience (pp. 47–63). https://doi.org/10.1007/978-3- https://doi.org/10.1016/S0950-7051(99)00020-9
319-74322-6_4
Mendling, J. (2009). Empirical Studies in Process Model
Verification. In van der A. W. M. P. Jensen K. (Ed.), Transactions
on Petri Nets and Other Models of Concurrency II (pp. 208–224).
Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-642-
00899-3_12
Muzi, C., Pufahl, L., Rossi, L., Weske, M., & Tiezzi, F. (2018).
Formalising BPMN Service Interaction Patterns (pp. 3–20).
https://doi.org/10.1007/978-3-030-02302-7_1
OMG. (2011). Business Process Model and Notation (BPMN),
Version 2.0. Object Management Group OMG.
Pack Kaelbling, L., Littman, M. L., Moore, A. W., & Hall, S.
(1996). Reinforcement Learning: A Survey. Journal of Artiicial
Intelligence Research (Vol. 4).
Polanyi, M. (1966). The Tacit Dimension (Edition 2009 with a new
forward by Amartya Sen). Chicago and London: The University of
Chicago Press.
Saif, H., He, Y., Fernandez, M., & Alani, H. (2014). Semantic
Patterns for Sentiment Analysis of Twitter. In The Semantic Web –
ISWC 2014 (pp. 324–340). Springer, Cham.
https://doi.org/10.1007/978-3-319-11915-1_21
Schallert, M., & Rosemann, M. (2012). Prozessreorganisation bei
einer Agentur für Unternehmens-, Finanz- und
Ressourcenplanung. In Prozessmanagement (pp. 579–597). Berlin,
Heidelberg: Springer Berlin Heidelberg.
https://doi.org/10.1007/978-3-642-33844-1_19
Silver, B. (2011). BPMN Method and Style, Second Edition
(Second Edi). Aptos, CA: Cody-Cassidy Press.
Soffer, P., Wand, Y., & Kaner, M. (2007). Semantic Analysis of
Flow Patterns in Business Process Modeling. In Business Process
Management (pp. 400–407). Berlin, Heidelberg: Springer Berlin
Heidelberg. https://doi.org/10.1007/978-3-540-75183-0_29
Staab, S., Erdmann, M., & Maedche, A. (2001). Engineering
Ontologies using Semantic Patterns. In IJCAI-01 Workshop on
Ontologies and Information Sharing . Seattle, USA.
van der Aalst, W. M. P. (2009). Process-Aware Information
Systems: Lessons to Be Learned from Process Mining (pp. 1–26).
Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-642-
00899-3_1
van Dongen, B., Dijkman, R., & Mendling, J. (2008). Measuring
Similarity between Business Process Models. In Proceedings of
the 20th international conference on Advanced Information