Beyond the Expressiveness of OWL for the Representation of Manufacturing Process

Tracking #: 3021-4235

Authors: 
Luis Ramos

Responsible editor: 
Guest Editors SW for Industrial Engineering 2022

Submission type: 
Full Paper
Abstract: 
Semantic Manufacturing is a trend that consists in the integration of Semantic Web technologies into manufacturing processes. This trend appears in the manufacturing domain because of the capabilities of knowledge representation and reasoning provided by Semantic Web technologies. However, Semantic Web technologies to date still exhibit shortcomings that are being inherited by the Semantic Manufacturing community. In this research a group of 18 ontologies, related to the manufacturing domain, were evaluated, following OntoSmart methodology1, determining from those selected that there were common issues among them. Some Competency Questions (CQ)s were compiled in order to use them as domain and quality reference. Moreover, we applied the name ‘Hypermodules’ to the common issues among pairs of ontologies, and we used the term ‘Hyperontology’ for the group of hypermodules. This Hyperontology was put into practice to represent the workflow of the manufacturing process. In this example, we showed that the Web Ontology Language (OWL) expressiveness for providing reasoning about such process was insufficient. We increased the expressiveness of OWL through the use of a Heterogeneous layer that included Distributed Ontology Language (DOL), Heterogeneous Tool Set (HETS) and Common Algebraic Specification Language (CASL). With the integration of a heterogeneous layer into our methodological framework, it was possible to model the most common constraints during the execution of such manufacturing process, as well as the validation of specific manufacturing operation configurations. Furthermore, it is worthy to mention that such validation considered the ‘hypermodules’ built from the evaluated ontologies, being this an evidence of ontology reutilization and extension within DOL. Finally, we were able to provide answer to 70 % of the CQ’s within this framework, which it would only achieve 23 % following other ontology development methodologies.
Full PDF Version: 
Tags: 
Reviewed

Decision/Status: 
Reject

Solicited Reviews:
Click to Expand/Collapse
Review #1
Anonymous submitted on 18/Feb/2022
Suggestion:
Reject
Review Comment:

The author has reviewed a number of ontologies developed in the manufacturing domain and applied the "OntoSmart" methodology to consolidate and evaluate them. While this type of work is appreciated and important for the community, the organization, presentation, and writing of the contents are simply ineffective and make reading of the paper extremely difficult. For instance, the introduction section has a lot of cursory discussion on digital manufacturing but does not highlight the actual work done or the contributions made (Also, the title is misleading and does not reflect the work done effectively). Shortcomings of existing work appear in the Methodology section (should be in the prior section) and the actual descriptions of the method are very shallow. In fact, most of such descriptions appear in the Results section, making it very hard to identify the actual results produced by the author versus the method applied. There are a lot of observations made in the Results section but there is a lack of flow and cohesiveness, which makes it impossible to identify the main points made by the author. Also in general, many sentences tend to run on with a lot of commas and should be restructured to be more concise and facilitate reading (e.g., pg:4, L:47-49). Because of these issues, it is impossible for me to provide any meaningful feedback on the actual work done. I'd suggest the author re-write the paper in a more organized and coherent form before it can be considered for a thorough review.

Aside: There are a number of important papers that used PSL (Process Specification Language) to model manufacturing processes. The author should review such work especially because they can overcome the modeling limitations of OWL. The authors should also look into and discuss Industry Ontology Foundry and their published work.

Review #2
Anonymous submitted on 14/Mar/2022
Suggestion:
Reject
Review Comment:

The paper proposes a methodology (OntoSmart) for developing and reusing ontologies. The author has evaluated 18 ontologies in manufacturing using the proposed methodology and has built a network of ontologies that can be used for answering a set of competency questions. Ontology reuse is a critical problem that needs to addressed particularly in the manufacturing domain so this paper is addressing and important problem.

It is clear that the author has enough knowledge of the field and a lot of work has been done in the context of this research. But because the selected scope of this paper is so wide and the challenges are so many, it is very difficult to address them all in a single paper. Unfortunately, this has resulted in a paper that suffers from lack of details and focused discussions in many places. The author is encouraged to reduce the scope and provide adequate discussion on the technical aspect.

Creating a hyper-ontology which is built through merging several ontologies is a very challenging task as the semantics of the classes (including the necessary and sufficient conditions) should be carefully evaluated to ensure there is no conflict among the imported classes. The paper doesn’t seem to provide adequate analysis on the semantic of the terms.
For example, the class ‘Product’ might have different semantics as the context changes. How the proposed methodology ensures that the class Product (as used in the hyper-module) is semantically consistent across all ontologies evaluated.

The title gives the reader the impression that paper is about OWL expressiveness issues with respect to manufacturing domain knowledge modeling but this is not the real focus of the paper. One major issue throughout the paper is that a lot of topics are discussed without providing adequate details about them. Therefore, it is very difficult to understand and evaluate the contributions of this paper.

Itemized comments:

1. The paper refers to General Information Quality Evaluation Questionnaire on page 8 (with a github link in the footnote). Is this a contribution of this research? If yes, it needs to be discussed in more details. In fact it makes more sense to write the entire paper around this single topic (ontology quality evaluation) since it is an important topic. How do you evaluate reproducibility, verifiability, Integrity, and timeliness? What does ‘timeliness’ mean for an ontology? A paper should be self-sufficient and requiring the reader to visit several external resources to understand the core concepts is not a good idea.

2. The proposed methodology, although comprehensive, is not a novel methodology by itself and the steps described for Ontosmart are the typical steps followed by most ontology development methods.

3. The paper is unnecessarily lengthy, and it is very difficult to follow some of the discussions.
4. Semantic Manufacturing is an odd term and, unless the author can provide a definition for it from the literature, it should be avoided.
5. Please don’t use the term ‘concept’ when referring to ontological classes. This is an ambiguous term. Instead, simply use ‘Class’.
6. The author claims that OWL does not have enough expressivity for representing manufacturing process knowledge but this claim is not supported by sufficient facts and discussions. What aspect of manufacturing process cannot be supported by OWL?
7. OWL can adequately represent parthood relationships for complex products (page 3). The author should discuss why OWL is not adequate for representing the structure of complex products.
8. How are you measuring reusability of the ontology ? what is the metric and how it is calculated? Detail really matter here.
9. There are some Competency Questions discussed in page 7. They seem very arbitrary. What is the basis for selecting these competency questions? CQs are ontology-specific. It is not reasonable to select the competency questions first and find an ontology that can answer those questions.
10. There are several algorithms mentioned in the text (with hyperlink provided as a footnote) but those algorithms are not discussed in the text.
11. What is domain strength? What is boundary strength ? how they are calculated?
12. What is Average of Defined Concept (AC)? How it is calculated?
13. The taxonomy shown in Figure 13 is semantically wrong. Unit is not a sub-class of measure? What is Prefix? How this class can be a sub-class of Measure ? what are the definition for those classes.
14. The reused classes from 18 selected ontologies can only answer 53% of competency questions. A properly designed ontology should be able to answer 100% of the CQs specified for the ontology. How the remaining CQ can be answered?
15. Is the application domain manufacturing or design? More clarity is needed

Review #3
By Michael Gruninger submitted on 10/May/2022
Suggestion:
Reject
Review Comment:

Technical Comments:

The Introduction does not provide a sufficient overview of the paper; in particular, it does not clearly state the key contributions of the paper. In a sense, I am also a little unclear about the exact nature of these contributions. The author is definitely presenting a new methodology for supporting ontology reuse. In this case, the methodology itself needs to be evaluated.
However, other possible contributions also seem to be implicit in the paper. For example, the design of a new ontology for manufacturing; in this case, the ontology itself should also be evaluated. Yes, the author mentions these points in the Introduction, but they need to be emphasized.

A third point raised in the paper (and the title!) is the question of the minimal expressiveness required to axiomatize the intended semantics of manufacturing processes. This is the least developed idea within the paper, and perhaps the most serious problem with the paper. Does the specification of the competency questions themselves require expressiveness beyond OWL? Does the solution of the competency questions require the additional expressiveness?
If so, what are the specific axioms that cannot be written in OWL?

These are general comments, bit more domain-specific comments can also be made. In particular, consider the competency questions in Section 4.3. Are these really sentences whose entailment require axioms within the ontology, or are they just lookup queries (e.g. SPARQL)?

The author also totally ignores any work in process ontologies (e.g. PSL, Event Calculus). This is particularly striking in the discussion in Section 5 (see lines 27-32 on page 22).

The work of Katsumi needs to be included.
I'm not saying that it solves the entire problem of reuse, and it definitely can be criticized from the perspective of usability in realistic applications. Nevertheless, it provides the formal foundations for many of the observations made by the author in this paper.

Other issues about reusability also need to be explicitly acknowledged. In what sense does searching for ontologies by keywords make sense?
Second, partial reuse -- is it possible to use a module or some other subtheory instead of the entire ontology if no single ontology provides answers to every competency question?
Third, is reuse by modification of existing ontologies better than design ab initio?
This question is alluded to at the end of section 4.4, but the discussion is very superficial.

I am puzzled by the notion of ontology quality presented in Section 4.4.2, and even more puzzled by a notion of ontology quality that is reducible to a quantitative score. This becomes an even bigger problem in Section 4.5, where there are insufficient arguments to justify the claims being made. For example, why is a number like Average Deployment of Concepts useful or even correct? Ontologies are primarily about axiomatizing intended semantics, and these proposed quantitative measure have nothing to do with semantics.

I question the utility of quantitative comparisons (e.g. Tables 5 and 6).

Editorial Comments:

Overall, the writing style can be improved.
There are too many paragraphs that consist of only one or two sentences.
Do these indicate that the ideas are not fully developed, or should some of these paragraphs be combined.

I can't parse the one sentence paragraph at lines 41-43 on page 35.

Should the title be
"Beyond the Expressiveness of OWL for the Representation of Manufacturing Proceeses"?

Recommendation:
The paper cannot be published in its current form.

The author should include more discussion on prior work on ontology reuse and process ontologies.
There is insufficient rigour within the paper.