A Critical Reflection on Ontologies and their Applications in Business

Tracking #: 706-1916

Wolfram Wöß
Christina Feilmayr
Nickolas Falkner

Responsible editor: 
Rinke Hoekstra

Submission type: 
Full Paper
In recent history, the term ontology has been used as if conveyed a great deal of weight and importance when, in many cases, the term has been used incorrectly. This diffusion of meaning is often the path by which a perfectly acceptable and well-defined word becomes a buzzword, reduced in meaning and a warning to readers that poor science is ahead. Frequently, a hyped buzzword will lead the reader to form expectations that are never fulfilled. This research work provides a critical reflec- tion on ontologies, their frequent misuse in research and business applications, and concerns aspects why ontologies have not been successful in large-scale business applications until now. When a definition that changes over time, as is the case for on- tologies, this may be indicative of a lack of understanding in the field, or an inability to effectively communicate and share a common understanding. Whatever the reason is for this case, introducing numerous definitions for one concept, especially a complex concept, leads to confusion and, consequently, people from various research communities can (and do) use the term ontology with different, partly incompatible meanings in mind. The general misuse of this semantic technology can be ex- plained by (i) the many existing, sometimes conflicting, definitions; (ii) too imprecise a specification of semantic technologies; and (iii) the existence of complex modeling processes that are too abstract or too complex. Because the term is used with a variety of meanings, only some of which are accurate, it has become increasingly difficult to discover who was or was not truly using ontologies. However, the relatively rare use in business applications can be attributed (i) to unknown (or unavailable knowledge about handling of) ontology modeling processes; (ii) to the lacking support of the modeling process because of missing state-of-the-art modeling tools; and (iii) to a lack of mature experience, developed over time. If we are to clearly demonstrate the benefits of an ontological approach, we must first clearly define what ontologies are and exactly when they are in use. Therefore, the causes for the general misuse and rare use of ontologies must be identified. Furthermore, it has to be clearly defined what an ontology is and guidelines are necessary to answer the question as to when ontologies should be used, how they can be used and when they should not be used. Finally, adequate representation languages should be applied to the problem and the ontology design process should be easy to understand and couched in terms that technical and non-technical users alike can understand. The main focus of this research work is to provide clear decisions to select a correct model, methodology and toolset to meet user requirements with the most efficient use of resources.
Full PDF Version: 


Solicited Reviews:
Click to Expand/Collapse
Review #1
By Victor de Boer submitted on 16/Feb/2015
Review Comment:

This paper aims to provide a critical reflection on the role ontologies play in business. The paper takes a critical view of the term "ontology" and its use as a buzz word. At the same time, throughout the paper, the lack of usage of ontologies in real-world business projects is identified and possible explanations are provided. The paper claims to aim to "provide clear decisions to select a correct model, methodology and toolset" for a given set of user requirements.

I found the paper very difficult to read and the main arguments and the claimed contributions of the paper change often in the paper (On page 1, 2, 3 ...) . There is a lot of repetition of claims, observations and suggestions, but none of these are presented in a coherent matter.
The paper is overly long and could easily be reduced to half its current length to get the same argument across. I here attempt to recreate this argument:

1. Ontologies were part of a hype
2. Adoption is very low. Many real-world cases which claim to use ontologies actually use taxonomies, structured vocabularies
3. This adaptation is low because:
- Businesses do not need sharing
- do not need the expressiveness of ontologies
- building ontologies is hard and costly
- Other methods suffice

Adhering to a structure like this would improve the paper considerably. In its current form, different parts of these claims are made in many different sections throughout the paper. For example, in section 8, where the title suggests that finally concrete suggestions for model, method, tool selections are discussed, the authors present again a bullet list of problems regarding current solutions. Such bullet lists occur frequently in the paper and no single coherent set of issues regarding the lack of adaptation is made.

Another problem is that for many of these lists, it is unclear how the authors got to this list. For example the list in section 4 ("From a technical point of view the usage of an ontology
can be of benefit for the following reasons: "). Where do these come from? Are they based on literature? original research?

While I agree with the overall point that fully-expressive ontology languages are often not necessary for certain tasks, the authors fail to specify which exact claims they are disputing in this paper. The fact that something is a buzz word is not very interesting from a scientific point of view. Disputing claims about usability or efficiency of using ontoloties is.

Other than these larger issues of readability and lack of coherence there are many issues regarding the paper. I here list only the larger ones and leave out smaller issues, typos, incorrect referencing etc.

p4: the point that taxonomies can only be used to increase recall in IR tasks is not true. I can easily think of examples where using a hierarchy can help in filtering out false positives.

p5: "The semantic web, by its goals and standardized nature, supports a strong middleware focus in the architecture but the presumption is that this is an automated middleware. " -> REference?

p6/7: "While ontologies (and taxonomies) are used within a large number of applications, it is rare for users to be directly exposed or required to interact with the ontology. " -> Why would users need to be exposed to the underlying technologies? I dont see how this is a sign of their usefullness?

p7: "RDF vs. OWL or, at its core, taxonomy vs. ontology." -> RDF is not a taxonomy language. SKOS is.

Section 3: The case studies from Section 3 are both overly long and underspecified. They are not case studies but observations in three different domains. Having three actual case studies would be interesting.

Section 3: In the second case study, the authors identify that ontologies are indeed "very useful" -which seems to be a counterexample to the main claim of the paper- but that there is no "one true ontology". It is unclear why there would need to be such a monolithic ontology or who claims that this should be there.

p17: the analogy on this page not helpful. If anything it is confusing.

fig4: The purpose of this image is unclear. The caption can be used to explain the elements.

p25: "But, there is no single API that travels with all ontologies." -> What do the authors mean by this? What would such an API look like.

Review #2
By Pascal Hitzler submitted on 18/May/2015
Review Comment:

EiC note: Paper has been withdrawn by the authors.