# Paper 87 (In-Use track)

Modeling Industrial Business Processes for Querying and Retrieving using OWL+SWRL

Author(s): Gabriel Silvatici Dayan, Suman Roy, Devaraja Holla Vaderahobli

Full text: submitted version

Abstract: Process modeling forms a core activity in many organizations in which different entities and stakeholders interact for smooth operation and management of enterprises. There have been few work on semantically labeling business processes using OWL-DL that formalize business process structure and query them. However, all these methods suffer from few limitations such as lack of a modular approach of ontology design, no provision for a consistent ontology development both with TBox and ABox axioms, no provision of combining control flow relations of the main process and sub-processes etc. In this work, we propose an approach for labeling and specifying business processes by using hybrid programs which offers modular ontology design, consistent ontology design of each module and unified control flow for process and its sub-processes. This formalism of hybrid programs integrates ontology specified in OWL-DL with SWRL rules. Further we report on our experimental effort on modeling industrial business processes with this hybrid formalism.

Keywords: Business processes modeling; patterns; ontology; OWL-DL; SWRL (Semantic Web Rules Language); querying

Decision: reject

Review 1 (by Daniel Garijo)

---------------AFTER REBUTTAL------------
I thank the authors for their answers. However, I have decided to keep my score:
* If real processes had been considered in the paper, then at least they should have been described on it. I suggest the authors use fragments of these processes in the examples to make them more compelling.
* The authors do not compare their system with related work.
* The authors do not assess the impact of the proposed approach.
---------------ORIGINAL REVIEW-----------
This paper expands a previous model for representing industrial business processes with a modularization approach and consistency rules. The authors evaluate their approach in terms of the time it takes to answer queries and analyze the satisfability of business processes.
The paper is well written, although sometimes the font size used in the examples provided, in particular in Tables 3 and 4, makes it a little hard to follow. The proposed topic is not novel, as there has been much work in business process validation with constraints in the past, but I think that the attempt made by the authors carries several lessons learned. However, my main concern is that the paper does not seem appropriate for the in-use track:
- The problem tackled is not supported by any real scenario to illustrate it (processes used in examples are process1, process2, which make the use case less compelling). The queries don't seem to be driven by requirements motivated by users.
- Evaluation: it is not very clear what the evaluation aims to measure. That the proposed approach is efficient? That is complete? that is useful for users? The time for answering queries and satisfability is shown, but the authors do not compare against any other work, just the size of the KB. The time needed to answer queries is not within the scope of the paper, right?
- There is no analysis on the impact of the approach (e.g., community usage, how the approach is used in a real world setting, etc)
- The contribution of the work is incremental, building heavily on a very specific piece of related work.
The SWRL rules shown in Figure 4 seem to be very specific for a type of processes. Have the authors thought about a framework to make the more generalizable? Are users supposed to encode these roles every time a new process needs to be represented?
I found a little weird that there are no mentions to BPEL (Business Process Execution Language)
Finally, some of the rules proposed seem a little complicated for the types of queries that are being answered by the system. Have the authors though about making the templates individuals and answer the queries using simple rdf/sparql according to the ontology? Shex/Shape rules could then be used to validate the business process.

Review 2 (by Alessio Gugliotta)

The paper presents a work of formalization of business processes in OWL-DL with SWRL rules for modeling and retrieval purposes. As properly discussed by the authors, this topic has been largely addressed by other past works, although some shortcoming still limit the wide application of semantic-based approaches in real-world applications.
One of the most relevant, current limits is the actual creation of semantic descriptions of business process models, which is a complex and error-prone task. The authors aim to address such an issue with an approach based on patterns, which, starting from a XML representation of BPM are opportunely combined, ensuring consistency. Then the patterns are also used to query created knowledge bases of business process models.
Although the use of ontology patterns to easily model domains and then query KBs is not novel (please refer to this research stream http://ontologydesignpatterns.org/wiki/WOP:2017), its application to business process modeling is indeed an interesting application.
Overall, the proposed approach is properly and clearly described, but additional details about the algorithm for generating composed BPM out of the basic patterns could be detailed.
In addition, and this is the main drawback of the present submission, the paper fails in demonstrating the actual benefits of using semantics in real world applications. The reported experimentations mainly focus on the querying aspects. The queried KBs contain descriptions of up to 29 BPMs, which is a quite limited number compared to the hundreds of BPMs used by companies for describing their business processes. And the reported experimentations also highlight that this is hard limit since "for KB with more than 40 process models, concept satisfiability for most of the cases do not terminate.". In other words the proposed approach cannot scale and it should be revised (ie. simplified) in order to deal with real applications.

Review 3 (by anonymous reviewer)

# Summary
In this work, the authors propose a way of encoding business processes in OWL to enable mining or querying business processes. The approach is not novel but it aims at overcoming limitations of the previous work. In particular, the authors show how to describe a process using more modular ontologies as well as how to describe subprocesses. Technically, they are describing well formed processes of most standard patterns of BPMN process language. The paper is finished with a proof-of-concept experiments.
# General Opinion
The paper is written clearly. Especially, motivation examples is  followed throughout the paper. And while the limitation of the previous work is clear, the main technical solution of this work is about replacing a complex TBox parts with new concept names (in fact, it is surprising that the previous work didn’t consider this). Still the authors provide further encoding advantages like combing OWL rules to enable encoding subprocesses. And while, the practicability of the whole approach is rather academic than practical, the paper looks as a solod work and ready to be presented.
# Some possible improvements/comments
- Up to my best knowledge OWL-DL relates to W3C standard that is deprecated and replaced OWL 2. See, https://www.w3.org/TR/owl-guide/. So OWL-DL is outdated for 9 years…
- I am not fully happy with detail explanation of motivation. Is there some evidence that using ontology reasoning (and encoding) has been adapted by somewhere (apart the motivating paper)?
- Querying processes is not the same as querying ontologies. In, particular, querying processes requires some other features like temporal dimension (using variants/extending LTL, CTL, etc). Could you compare your approach with other (I would say more academically adapted) approaches?
- Do we need such complex ontologies. As the authors observe querying with such complex ontology (NEXTPIME) is very expensive. Could you identify some either more simple fragment of OWL or use one  of the OWL2 profiles (https://www.w3.org/TR/owl2-profiles/)
# Minor Comments
- In first paragraph Sec 1: “business process models” are I assume different process model languages. Still, later the same phrase is used for a set of concrete processes (e.g., Maintenance Processes). Please clarify the terminolgy.
- In general latex formatting the paper is poor. Eg., $M P rocess1$ should be better formatted, e.g.,  $\textit{MP rocess1}$. Similarly, the same problem elsewhere.
# After Rebuttal
I thank the authors for their response. Based on the response and the discussion among reviewers, I now more of the opinion that the paper not a proper in-use paper since it lacks real-scenario motivation and impact assessment (which I overlooked initially). Considering that it also technically shallow, I decided to downgrade my total score to reject.

Review 4 (by Jacco van Ossenbruggen)

The paper describes a conceptual mapping from Business Process Modeling Notation (BPMN) into OWL (both A-Box and T-Box), an implementation and evaluation section.
Without going deeper into the content of the paper, I think the paper does not fit the scope of the in use track.  While the research was carried out during an industrial internship, the paper does not make clear which "real" industrial problem is being solved and how the presented solution could impact the work done in the company. Also, there is no indication that the solution has actually been in use or that there are plans to put it in use in the future. Also, in the evaluation, it remains unclear how the test query set is representative for the problems at the company. At best, this is a report on an internship project that is using company-data.

Review 5 (by Anna Tordai)

This is a metareview for the paper that summarizes the opinions of the individual reviewers.
The reviewers mention that the paper is well written and that the approach is clear. They point out that the paper does not include a description of real use or an analysis of the impact of the work and therefore is not appropriate for the In-Use track. Additional comparisons to related work would improve the paper.
Laura Hollink & Anna Tordai

Share on