Future Internet Ontologies: The NOVI Experience

SWJ review

The paper discusses the application of ontologies on Future Internet platforms, it presents specific ontologies and it examines how different models of ontology mapping (direct, indirect, hybrid) perform.

The positive aspect of this paper is that it is based on research work that has been performed under the auspices of a collaborative research project. The negative side is that its contribution is not very clear as the paper stands and it is not rigorously defended. In particular, there are a number of issues that need to be addressed before its contribution can be established:

There are a number of issues when trying to evaluate ontologies. If the resource, monitoring and policy ontologies in the context of FI are the main contribution of this paper, there need to be references on ontology evaluation and a detailed description of not just the ontologies but also the methodology and process of their evaluation. At the moment, there is some discussion (not exactly evaluation) in section 7.1 and a brief outline of the ontologies in section 3.
Some of the challenges that are particular to FI when applying ontologies need to be better introduced and justified. It is stated in the abstract that FI ontologies continually evolve but it is not discussed later in the paper whether this is particular to FI or to ontologies in general?
Claims like the one on federation in the third paragraph of section 1 can be contentious. They need to be rigorously discussed especially if they are central to the motivation of the research work.the same applies to the content of the fourth paragraph of section 1; some examples of how deduction would help FI in particular would be helpful.
The challenges towards the end of section 1 could be more clearly articulated and more factual. For example, how do we know that the mapping of concepts it the ontology is an “important” decision that requires careful consideration? Why? What is at stake if mapping of concepts is not carefully considered?
In the beginning of section 2 it is claimed that a “thorough” review focused on a number of models, which are detailed in the following paragraphs; how do we know that the review was thorough? what process was followed? If a minimal part of those ontologies could be used then what was the research challenge of providing new ontologies?
Section 3 tends to be laconic in outlining the NOVI ontologies without much elaboration. If those are provided already in the project deliverable, what is the point of repeating them in the paper? How does it help establish the paper’s contribution?
Similarly, what is the purpose of section 4? If the architecture is a contribution of the paper then what have been the challenges that it is trying to address?
In the beginning of section 5 it is stated that a number of lessons were learned but there is no detail of how those lessons were learned or argumentation on why/whether they apply in different contexts.
In section 5.3, why is the hybrid mapping not a special case of the indirect model mapping?
The evaluations of section 7 is more of a brief discussion, while the recommendations of section 8 are not rigorously argued; intuitively, they appear to make sense but how did the authors arrive at those contributions? is the process reproducible?

A couple of minor points:

NOVI as an acronym needs to be explained and citations, URLs, contract numbers would be helpful.
The English of the paper could be improved.

Summary of review:
In this paper, the authors proposed an information model via some ontologies in the area of Future Internet. The project is called NOVI, and its goal is to define some ontologies in order to support some features such as context awareness, monitoring, management policies and virtualization. The paper states that the realization of their goals is obtained by proposing three main NOVI ontologies: NOVI Resource Ontology, NOVI Monitoring Ontology and NOVI Policy Ontology.
The paper explains two basic approaches for mapping information model to object model. In fact, the approaches discuss how to use ontologies to create object models for software components in the Future Internet area. The two main approaches are direct object model mapping and indirect object model mapping. In direct object model, there is a direct mapping between concepts of ontologies and classes of object models, and keeps the programmers away from dealing with knowledge of domain. NOVI uses Alibaba tools to provide direct mapping from Java objects to and from RDF resources, and OWL classes to Java classes. However, in indirect object model, the structure of object model is not exactly the same as the structure of information model, and only considers the conceptual relationships.
Finally, they describe how these two mapping methods can confront with some challenges in software development such as ontology changes, software scalability and domain changes.
There are following weaknesses in the representation, contents and evaluation of the approach:
• The paper starts with an important motivation that ontologies allow deduction of services and infrastructures. So, we can benefit from these features in the area of Future Internet especially in policy and monitoring data. Moreover, authors compared their model with some existing models such as CIM, and explain that CIM could not provide support for semantic, because it is based on UML. This is an obvious fact that UML is semantically weaker than ontologies, and ontologies have much more power to explicitly represent the conceptualization. This power is valuable when we can use them in order to have computational and logical inferences. Reasoning is one of the most important capabilities of the ontologies that we can use because of their logical backbone. Unfortunately, we cannot find sensible reasons and benefits of using the potential of ontologies in this paper. We cannot find the issues that UML cannot deal with in CIM and ontologies can in NOVI. In section 8, there is a sentence: “we do not have enough data to say if there is a clear advantage of using OWL ontologies in FI description and management”. This can be a serious defect for the paper, and weaken the motivation and goals of the work.
• The title of section 4 is ‘The NOVI architecture and its use of the ontologies’. This part describes the NOVI architecture, and shows an example as a user unbound request expressed in Turtle. There is nothing special for the use of ontologies in the architecture; there is only one example of how the user request is expressed via Turtle and ontologies. Definitely, we can have better usage of ontologies in the architecture that have not been mentioned here.
• Evaluation part of the paper is not actually an evaluation part, and only indicates the effects of direct and indirect mapping strategies on the three mentioned challenges: ontology changes, software scalability and domain changes. The effects of using ontologies on finding and deducing appropriate services can be considered as an important aspect of the evaluation.
• We can find detailed information about the different parts of NOVI ontology and its implementation in [1]; however, there is not enough information about these parts in the paper.
• We have noticed a significant overlap (one of the authors overlaps: Cees de Laat) between the submitted paper and [1] which is a recently published journal paper and has much more complete and with detailed information. Given that the submission is also a journal paper (not a conference/workshop paper submitted in advance of the current submission) we would like to have a thorough description of the differences between the two journal papers.

Minor comments:
• 3.2. NOVI Monitoring Ontology: defines four groups of concepts in section 3.2, and Fig 3 also shows the classes of this ontology, but there is no explanation about the classes of this ontology in the figure, and their relation to the specified four groups of concepts.
• 3.3. NOVI Policy Ontology: this part describes the operation of Policy Service. We cannot find this service in Fig. 2 as a subclass of Service.
• 4. The NOVI architecture and its use of the ontologies: Fig. 5. Shows the architecture of NOVI, and its description is explained in section4. However, there is a component in the text (IRM: Intelligent Resource Mapping) that we cannot find its corresponding element in the figure and its relations with other components.
• 8. Recommendation for FI adopting ontologies: authors provide four recommendations for other projects in the area of Future Internet, but they left this part without any explanation and justification about their claims.
• 5.1 an ontologies -> an ontology

[1] Jeroen van der Ham, József Stéger, Sándor Laki, Yiannos Kryftis, Vasilis Maglaris, Cees de Laat, The NOVI information models, Future Generation Computer Systems

This paper provides the high level description about the NOVI ontologies, but it lacks of the real data testing, and real use scenarios, and real deployment. It is not clear why we need such ontologies, how to deploy such ontologies, why this approach is better than other non-ontological driven approach.

Introduction
Yes, ontology plays an important rule to coordinate and integrate difference resources. But creating, maintaining, or updating an ontology requires tremendous efforts, especially when these ontologies have more in-depth logical functions. There are few issues for this approach: 1) what will be the efficient way to create and maintain these ontologies; 2) if you need several ontologies, how to align them or map them will become the bottleneck for further applications; 3) how to deploy such ontologies, if given your infrastructure is distributed, there is no central controlled place to enforce such ontologies.

What is the difference between NOVI ontologies with the OWL-S ontology?

Also, this paper lacks of the design requirement or rational why you design your ontologies in such ways. What is your environment, what are the requirements your ontologies need to fulfill?

evaluation and scalability are not discussed in depth and tested in real data or real environment. How to compare your approach with existing similar approaches, what are the pros and cons? How to handle large scale data? What is the scalability of Alibaba, Jena and sesame cannot handle large scale data?

How do you handle the URL or create URL/URI for your objects? What is your naming convention, how do you deal with the same objects with different URI? Also, how do you extract objects from existing non rdf but full-text service descriptions? or you assume that people will use your ontologies to describe their services, and ignore existing services which were not modeled using your ontologies.

Summary of the paper

The paper ontologies, which have been developed the European NOVI project,
for Future Internet (FI) applications.
The motivation is to use ontologies (and semantic web technologies in general)
to meet the challenge of managing federated heterogeneous resource on the Web,
as required in the FI context. Besides the presented ontologies (NOVI ontologies),
the paper describes how the NOVI architecture uses these ontologies,
including reports on experiences and lessons learned.

Evaluation of the paper

The idea of using ontologies to cope with the management of distributed and heterogeneous
resources is quite meaningful and promising. The potential of semantic web technology
in the realm of FI is outlined. It is promising to use the developed ontologies as a
mediator between NOVI services and the physical layer. It is positive that a lot of work
and implementation has been done within the NOVI project. The mapping descriptions (Sect. 6)
and the recommendations (Sect. 8) are useful, even if it is written at a very high abstraction level.

Unfortunately, the paper has several weaknesses.

First, the presentation is very bad. Besides typos and grammatical errors (see below)
the storyline of the paper is not very clear.
The paper reads like a summary of project results,
but not like a research paper. The ontologies are mentioned and the NOVI architecture is introduced,
but none of them is properly explained. It is not clear why ontologies are
needed and how they are used in the system.
The overall motivation is missing and a coherent story can not be found.

Second, the need for a new modeling approach (i.e., the NOVI ontologies)
is not well explained. Sect. 2 mentioned features like virtualization,
monitoring and policies, but it is neither stated where these features / requirements
come from nor why ontologies are needed.

Third, again the bad presentation makes it hard to understand the modeling decisions.
In Sect. 3, the basic ontologies are introduced. It is not described why three parts
are needed and there modeling scope is mentioned, but not explained, i.e., it is not
clear what “Node”, “NodeComponent” and “Network Element” are used for and where these
entities are originally defined.

Fourth, the architecture in Sect. 4 is not understandable due to limited explanation.
It is not clear where and how the ontologies come into play.

Fifth, the most important criticism is the missing explanation on ontology modeling,
knowledge engineering and application of semantic web technologies like reasoning
and querying. There is only a short excerpt of some request specified by some
triples. However, any interesting experiences and modeling issues for
semantic Web interested reader is not provided.

Sixth, related to the previous issue, the experience description (e.g.,
regarding the mapping principles) could be some interesting work for semantic
Web researcher. However, there are only very few details given and the
description is mainly on a very high abstraction level,
discussing rather known issues on triple-object mapping.

Minor comments
- Several grammatical errors (especially various commas are missing)
- Figures (e.g., Fig. 3) are not readable
- Several typos, e.g.,
+ p. 6 these model -> this model
+ P. 6 object model are -> object models are
+ Alibaba vs. AliBaba
+ P.9 a different implications -> a different implication
+ P. 10 to is what -> to what