Review Comment:
This paper describes the semantic extensions to an IoT application development tool called Node-RED. Semantic application templates called Recipes are used for matching the skills or the capabilities of things with the application requirements. Models from iot.schema.org are used to semantically enrich the nodes from Node-RED. W3C WoT Things Description (TD) are generated automatically from the iotschema nodes of Node-RED. These TDs are stored in the form of a Knowledge Graph either centrally on the Cloud or distributed across the Edge devices. Experiments were conducted using both these approaches and results were discussed.
I have the following concerns with the papers
1) I think this should be an "Application Report" rather than a full paper. I am not sure what the research contributions are. It is not clear from the paper and neither are they specified explicitly. It is a very nice application of Semantic Web technologies in solving problems in the IoT space. But beyond the standard modeling and querying techniques, I fail to see what the research contributions are.
2) The following information is spread across the paper and is not completely clear - what are the specific additions and work that is different in the journal submission as compared to the authors' earlier publications? Please put this information in the Introduction section (towards the end perhaps) and mention clearly what the add-ons are.
3) What parts of the entire workflow (for example from Figure 1) are automated and which of them are not? It looks like only the matchmaking part is automated.
Other questions/comments
a) When explaining the approach in Sections 4, 5, 6, and 7, it will be good if authors can map it back to the development steps from Figure 1.
b) The recipe flows and matchmaking in Node-RED seem related to Semantic Web service composition and discovery. Please include relevant papers in the related work and comment on how this work compares to the composition and discovery work from the past.
c) The term semantic reasoning has been sprinkled across the paper but it is neither explained nor clear how reasoning is used and what reasoner is used (is it VLog?).
d) The details of matchmaking are not clear. From Section 7.2, how is Recipe Flow converted to SPARQL queries? How do SPARQL queries help in matchmaking? Is matchmaking always a binary case (match or no match) or will be there cases of "closest" match? For the use case in the paper, mention how SPARQL queries are generated and also put the queries in the paper.
e) The phrase rapid application development has been used throughout the paper but the term "rapid" has not been defined or quantified by the authors. How rapid is rapid? As part of the user studies I was expecting the following - time taken by the users before and after the semantic extensions to Node-RED. This is missing and the authors claim this to be the key selling point of their work.
f) How does the TD Generator semantically enrich TDs from iotschema nodes (Section 6.1)?
g) Did you try the standard triple stores on Edge devices? How was that experience (performance, memory usage etc.)? Putting these things in the paper would serve as a better motivation for choosing VLog.
h) Since the VLog store is distributed across the Edge devices (Section 7.2.2), how much of a communication overhead will there be while loading TDs and querying them? Do you meant "replicated" instead of "distributed", i.e., all the TDs are replicated on all the Edge devices (and not distributed across them). If it is distributed then my question on communication overhead holds. If it is not, then the terminology used in the paper has to be changed.
i) Page 23, Q1, common practice is mentioned here but it would be helpful if the common practice was discussed early in the paper (so that the readers can appreciate what the difficulties in using non-semantic Node-RED are).
j) Page 8, the format of text on the left side is different (line spacing is more?) from rest of the paper.
k) The quality of some of the figures is not good - Fig 3, 4, 5, 6 for example.
l) Fig 5 needs a legend for the different colors used.
m) Section 5, what are brown-field and green-field things?
n) Section 5, can new adaptation nodes be created by the users?
o) Page 22, under "User Evaluation with Siemens Engineers", the sentence "... user evaluation of the tool with Siemens expert engineers (who are non-experts)." is not clear. Who are the experts (and in what) and who are the non-experts (and what are they non-experts in)?
Spelling/Bibliography/Grammar corrections
i) Page 1, it should be digitize and digitization rather than digitalize and digitalization.
ii) Page 2, "... and expensive. As the state of the art IAS and BAS are engineered ...". Starting with As is not correct and is abrupt - there is not continuation from the previous sentence. This needs to be rewritten.
iii) Page 7, in the sentence "... maximum one input and one ore more ...", it should be or and not ore.
iv) Page 23, summary is misspelled as summery.
v) Some of the references are missing year and/or pages. Reference 28 has weird characters in it.
vi) Page 22, it should be "On" instead of "At" in the sentence "At this occasion, ...".
Overall, this work is a very good demonstration of how Semantic Web technologies can play a key role in making IoT application development better and easy. However, as mentioned above, the research contributions and some parts of the paper are not clear.
|