Graphical Transformations of OWL Ontology to Event-B Formal Models

Tracking #: 2209-3422

Eman Alkhammash

Responsible editor: 
Rafael Goncalves

Submission type: 
Full Paper
Formal methods use mathematical models for developing systems. The creation of formal models from informal requirements demands skills and effort since this involves major problems such as ambiguity, inconsistency, and imprecision. To tackle this, it is necessary to have methods and approaches to support mapping requirements of formal specifications. This paper aims to present an approach that addresses this challenge by using the W3C Web Ontology Language (OWL) to construct Event-B formal models. OWL ontologies provide the formal notion that capture domain models. Recently, research on requirements engineering have shown an increased interest in using ontologies in various phases of requirements engineering. Our approach reduces the burden of dealing with formal notations of OWL ontologies and Event-B models and aims to harness the power of OWL ontologies to construct Event-B models using diagrams. The idea is based on the transformation of OntoGraf diagrams of OWL ontologies to UML-B diagrams to bridge the gap between OWL ontologies and Event-B models. OntoGraf is used to visualise ontology knowledge, whereas UML-B provides diagrammatic modelling notations for creating Event-B models. Event-B supports stepwise refinement to allow each requirement to be introduced at the most appropriate stage in the development in order to manage complexity. UML-B supports refinement and, therefore, we also introduce an approach that allows to divide and layer OntoGraf diagrams into a number of layers that can be introduced as refinement in the UML-B diagrams.
Full PDF Version: 


Solicited Reviews:
Click to Expand/Collapse
Review #1
By Dmitry Mouromtsev1 submitted on 08/Oct/2019
Review Comment:

The paper presents an approach for graphical transformation of OWL-ontologies to Event-B Formal Models. The idea of diagrammatic modeling and reasoning is quite natural for ontology engineering and this task lacks efficient methods and tools. At the same time the problem itself has many issues that depend on a specific modeling task and a domain. So, a research of graphical representation of OWL ontologies should have deep study of many aspects: graphical grammar for complex semantics of OWL, rules of data transformation for visual representation, understandability of ontology diagrams, user study and usability and many others. In this sense a contribution of the paper is not clear, and it require a lot of work to be done to be acceptable for publication.
The most critical points for rejection are following:
1) The author selected OntoGraph as reference tool for their work, but it is not clear why namely this tool was chosen? There are numerous OWL visualization tools and methods and a survey of related works could explain the state of the art and author's motivation. A good review of ontology visualization methods and tools can be found in [0]. The given related work description in the section 7 does not answer this question.
2) The paper is overloaded of obvious staff about ontologies and UML diagrams. It may be skipped without any harm for understanding of the main idea.
3) In spite of given case study, the paper does not contain any description of datasets that were used during research for construction of visual representations. A more detailed analyses of such a data could help to prove a correctness of the proposed method.
4) From the method description it is hard to understand what is its benefits? For example, from the figures 6, 7 and 8 it is not possible to understand the benefits of a new method. And the corresponding text provides only very few details.
5) An evaluation of the proposed method is not well designed. The only an application of the proposed approach to a couple of case studies is given. But these case studies are vague and just describes the built diagrams but not prove the method advantages.
6) In the section 6 there are quite obvious points without any deep discussion of the research results.
7) Finally, the conclusion section declares a number of very important problems, but the paper does not contain enough information to conclude that these problems are solved in the presented research work.

Unfortunately, I can't recommend to accept the paper for publication in the current state. I suggest to figure out more specific research problem and to carry out its evaluation deeply.

[0] Dudáš, Marek & Lohmann, Steffen & Svátek, Vojtěch & Pavlov, Dmitry. (2018). Ontology visualization methods and tools: a survey of the state of the art. The Knowledge Engineering Review. 33. 10.1017/S0269888918000073.

Review #2
By Vojtěch Svátek submitted on 14/Nov/2019
Review Comment:

The paper addresses an important area: the use of ontologies as support in requirement engineering.

The overall idea presented by the authors (if I understand it right) is appealing: developing an ontology, tuning it with respect to logical consistency, dividing it into multiple layers within a graphical ontology editing tool, and the automatically converting it into a layered Event-B model that would enter the software development process as a domain model.

I however have several principal problems with this paper:
- No scientific challenge is described. The examples are rather trivial. The way the authors serve the transformation process, it seems that the description logic constructs of OWL are simply translated to their first-order equivalents.
- The explanation of Event-B and UML-B is too laconic and fails to grasp the core of the approach, for a person unfamiliar with those languages (which many of the SWJ readers would be). At least, the concrete diagrams and code snippets should have been thoroughly explained.
- One of the crucial steps, the decision about the layering, is not properly demonstrated. The authors only write: “we decided…” But why this way and not some other?
- There are some misunderstandings on the authors’ side as regards the semantics of OWL.
- The quality of English is seriously compromised. There are typos, grammatical errors and ill-formed sentences all around. Also the typographical quality is not good, e.g., the use of screenshots for displaying code is inappropriate. Some parts of the paper only consist of headings and floating figures. Even some basic identification info is missing, such as the country of the authors, in the affiliation. Some bibitems (such as [11] and [20]) are very incomplete, and some contain funny characters.

Some detailed comments:
- The XML notation in Fig. 1 is not very pleasant to read.
- Examples of events in p. 3 are not very clear. For example, does R always have the same meaning (replacement?) despite the varying arity?
- In Fig. 3 I do not understand why “member” appears both as an attribute and an association.
- Some entities of OWL ontologies are mentioned to be transformed to “invariants or axioms”. However, there is no explanation on what basis one or the other is chosen, and there is no example of axiom generation shown.
- “The property can be of ”inverse type”.” Here the authors probably failed to grasp the meaning of “inverse functional property”. This is not “a functional property that is of some “inverse type””, but “a property the inverse of which is functional”…
- The first para of Section 4 already speaks about the LUBM ontology entities, this is clearly a mistake.
- In Section 5.1: “…affiilatedOf, head, and affilatedOrganization are represented as partial functions…” The notion of partial function is not mentioned anywhere else in the paper.
- The PO acronym is used many times but is never explained. I actually cannot find how to expand it… “proof… something”?
- “…are generated from m0, m1, m2 and m3 Event-B models and automatically discharged.” What is meant by “discharged”?
- The situation concerning the weather ontology, that 8 of the proofs (?) in M1 had to be interactive (?), would be worth explaining.
- Why so much focus on OntoGraf? Wouldn’t any graphical tool for OWL editing serve equally well, if the point is just to be able to arrange the nodes graphically so as to clearly see the layers as separated?
- “This definition brings us to think of modeling events based on object properties and data properties of OWL ontologies. For instance, we can introduce the eventgetMastersDegreeFrom from the object propertymastersDegreeFrom in Figure 31” The interplay between an event and the relationship it establishing is an interesting topic. However, the authors do not explain how to automatically recognize that there is some event behind a relationship (or vice versa).
- “Another work by [22] aims to build Event-B models from OWL ontologies. It uses OWL-verbalizer to generate controlled English requirements called Attempto Controlled English (ACE) from OWL ontologies which can be then used to construct Event-B models.” This is earlier work by the same author. However, since ACE can be directly translated to OWL, it is actually very similar work to the current paper. The delta of the two should be better explained.

Assessing the paper by the standard major axes: 1) its originality might be fine, but I cannot judge it properly since I am not thoroughly familiar with the applications of ontologies in software engineering; 2) the significance of the results in the current stage is poor – the demonstrated results mostly look trivial; 3) the writing quality is poor.

In all, in the current form the paper is unfortunately far below the SWJ standards. I suggest that the authors might try to rewrite the paper focusing on the interesting scientific challenges, provided there are some. They should also pay more attention to the formal treatment of the relationship between OWL models and Event-B models as logical theories, and, significantly improve the writing in all respects.