Reviewer has chosen not to be Anonymous

Overall Impression: Good

Content:
Technical Quality of the paper: Good
Originality of the paper: Yes
Adequacy of the bibliography: Yes

Presentation:
Adequacy of the abstract: Yes
Introduction: background and motivation: Good
Organization of the paper: Satisfactory
Level of English: Satisfactory
Overall presentation: Good

Detailed Comments:

The paper has a clear and well-stated purpose: developing a neuro-symbolic system for automated IoT platform configuration. By leveraging the complementary strengths of symbolic and connectionist AI, the paper proposes a system that streamlines the configuration process of IoT platforms by following a design science research approach, introducing two key artifacts: the Instantiator Pipeline and the IoT2Model method.

It is a well-structured and well-written paper, where the three main dimensions of the work are well-stated, introducing the fundamentals of IoT, neuro-symbolic AI, and the development of domain-specific modeling methods. Afterward, problem context, artifact, design requirements (derived from a related work analysis), and goal are delimited following the Design Science research methodology that the paper follows. In the author's words, "improving the efficiency of IoT platform configuration" constitutes the problem context, "by developing a neuro-symbolic AI-based pipeline" fixes the artifact, " to enable device discovery, platform configuration, openness, scalability, platform interchangeability, and model-based scenario specification" refer to the relevant requirements, and "to provide a holistic system for IoT environment configuration" is the final goal.

Globally speaking, the paper fulfills its main goal of developing a design science artifact that enables the automated configuration of IoT environments. It presents first the specification of the design artifact, namely the Instantiator Pipeline, which operationalizes the conceptual framework outlined above, before a concrete configuration case based on this implementation is showcased. Additionally, the implications of the Model-based Scenario Specification requirement (R6+) are discussed,and the use of the Instantiator Pipeline is presented within a basic experiment environment setup of a Smart House scenario.

Summarizing, the paper does a good job in providing all the discussed results. It contributes to developing a comprehensive design artifact that intends to automate and enhance IoT platform configuration by integrating neuro-symbolic AI with conceptual modeling techniques. Said that, some aspects could improve the clarity and intended quality and value of the contribution. They are suggested below:

- The goal of answering how neuro-symbolic AI and conceptual modeling can be applied to enhance the configuration and customization of IoT platforms is a really interesting problem. But what AI perspective does the paper take? The questions of what neuro-symbolic AI techniques are used and how AI and CM are in particular connected have not the clear answer that a reader who is attracted by the interest of the paper topic expects. i

-Following up on the AI use argument, what is the precise use of LLMs? How do they exactly contribute to the presented approach?

- The process of creating scenario-specific rules using the IoT2Model tool in the context of the Smart House case appears to be too generic and too vague in concrete practical terms. It would clearly benefit from adding some more precise practical details. From this practical perspective, the description becomes too unclear in terms of easily understanding the intended benefits of the solution, and what is improved with respect to what can be done with currently existing approaches.

- What benefits are derived from the use of the ADOxx platform? Why ADOxx? How well did it adapt to the system's requirements for development?

Reviewer has chosen to be Anonymous

Overall Impression: Good

Content:
Technical Quality of the paper: Good
Originality of the paper: Yes, but limited
Adequacy of the bibliography: Yes, but see detailed comments

Presentation:
Adequacy of the abstract: Yes
Introduction: background and motivation: Good
Organization of the paper: Needs improvement
Level of English: Satisfactory
Overall presentation: Good

Detailed Comments:

The article presents activities related to the development of a Neuro-Symbolic approach for automating IoT device discovery and IoT platform configuration.

I must admit the topic is really interesting despite some flows emerge in the current version.

Nothing is mentioned on possible limitation of the approach. E.g., are Knowledge Graphs of IoT platforms already available? Is that limiting the applicability of the approach?

In the article references to the concept of Digital Twin are reported but the concept is not introduced. Especially consider that DT is a controversial topic that has been attributes several different definitions. The authors should clarify their stand reporting the DT definition they refer to for this paper.

In addition it is not clear why the DT topic is mentioned at all.

It is also not clear why the IoT platform configuration should automatically mimic the physical environment and why IoT devices should be removed from the IoT platform. Why is that relevant?

It is not clear what happens after a device is deleted from the IoT platform. Are all data (also historical data) removed?

It is not clear to me whether ADOxx is open-source. Open-source it has been claimed for all the other components but nothing is mentioned about IoT2Model. Is the IoT2Model open source?

In the approach some assumptions are made such as: "Each of the sensors publishes data to a dedicated topic, allowing subscribers to receive real-time updates." Those assumptions should be better emphasised together with possible limitation of the approach.

Approaches are available in the literature that provide metamodel and support for designing IoT application to be deployed on IoT platform. I notice some similarities to what the authors are presenting here and might be worth a further inspection. I list those that come to mind:
- Corradini, F., Fedeli, A., Fornari, F., Polini, A., Re, B., & Ruschioni, L. (2023). X-IoT: a model-driven approach to support IoT application portability across IoT platforms. Computing, 105(9), 1981-2005.
- Corradini, F., Fedeli, A., Fornari, F., Polini, A., & Re, B. (2022, October). DTMN a modelling notation for digital twins. In International Conference on Enterprise Design, Operations, and Computing (pp. 63-78). Cham: Springer International Publishing.

Section 5 is a bit off. Especially the first paragraph "The discussion aims to both motivate the extension of our approach through IoT2Model and highlight future work regarding the benefits of its utilization within educational settings. The motivation for extending our approach with the IoT2Model method " makes a reader think that the work is not original but it has already been reported elsewhere and what is added in this article is just the IoT2Model. I am confident that this is not the case but in anycase the section should be revised. If the article is an extension that should be stated in the introduction section.

In addition anyone familiar with IoT platform knows that they not only support the registration of devices.

It is not entirely clear to me where the R6+ comes from.

Authors claim that the approach simplifies complex configuration processes. It is not clear whether the approach has been validated therefore it is not clear on what basis the authors can claim that.

I recommend a careful revision before a possible publication.

Minor:
- almost all the quotation mark present issues. In latex the opening mark is different from the closing one. Please fix them everywhere
- Fig. 3. Sniff Loop exit arrow has label "false" but it is not clear the meaning of that.
- Fig. 6 "Atleast two states" --> " At least two states"

Reviewer has chosen not to be Anonymous

Overall Impression: Good

Content:
Technical Quality of the paper: Good
Originality of the paper: Yes
Adequacy of the bibliography: Yes

Presentation:
Adequacy of the abstract: Yes
Introduction: background and motivation: Good
Organization of the paper: Satisfactory
Level of English: Satisfactory
Overall presentation: Excellent

Detailed Comments:

The paper at hand uses a neuro-symbolic approach for matching IoT device data with a knowledge graph in a Neuro-Symbolic (NeSy) transformer, such that configuration information for IoT platforms can be identified, devices can connect to the platforms and become discoverable, and be applied in scenario-specific applications. Methodologically, design science is used to establish the motivation, requirements, artifact, prototyping, and overall process, the artifact implementation utilizes a sentence transformer model (paraphrase-MiniLM-L6-v2) for matching, and conceptual modeling is applied for the scenario-specific applications.

Strengths:
- The main contribution is linking conceptual scenarios and applying them to a combination of IoT data matching with a transformer and knowledge graph. In this way, the transformer and graph are utilized for automation, while scenarios can be defined on a conceptual level. The contribution is, therefore, a holistic approach of applying the Neuro-Symbolic transformer approach. This is, clearly, a relevant and novel contribution.
- Overall, the methodology is well defined and applied, including the overall method, technical components, and conceptual components.
- The paper balances these aspects and manages to also present relevant technical details.
- Examples and links to the implementation are available.
- It is a well-written and well-structured paper.

Minor weaknesses:
- Sections 1, 3.1 and 3.2 motivate and introduce relevant concepts, however, the high-level explanations could be shortened to define the concrete research problem (3.1) earlier.
- The problem itself should focus more on applying conceptual scenarios with Neuro-Symbolic transformers, since this is the actual content of the paper and the topic of the special issue.
- The DSRM template is given twice on pages 4 and 6 in slightly different forms. For clarity, it should be presented once.
- Section 4 could be structured in a way that follows the structure of the proposed solution. E.g., the first part of the architecture (initiator platform) is introduced, which could be reflected in the title of the section or a subsection. The section title "Automating the Configuration [...]" is concerned with the use case.
- KG data is specific to the scenarios and devices, e.g., involving sensors and actuators. Devices in various industries might employ different processing, data storage and retrieval and connection technologies that might not produce meaningful matches for the NeSy-transformer. The paraphrase-MiniLM-L6-v2 transformer might find the most similar configuration data and it applies it, which might not be optimal. That is to say, the main contribution is the transformer-based matching with the knowledge graph. The specificity should be clearly pointed out, possibly highlighting the benefits of matching to provide exact data.
- As a result of the instantiator pipeline, devices are discoverable and configured for an IoT platform. The paper also mentions "digital twins" that are configured on the platform through this configuration. However, it does not go into detail and it remains unclear how the configuration, e.g. the presence of device sensors, would be sufficient to create a digital twin. For example, the presence of a device sensor does not specify its technical measurement parameters, value range, voltages or other information relevant to the operation of the sensor or even the interpretation of its results. Either the extent of the creating digital twins need to be clarified, or the few mentions of it should be removed.
- The paper could stress the scenario-specific application further that makes it a holistic approach. In this way, rules implemented through conceptual modeling allow for individual and specific configurations and the NeSy-approach can be applied meaningfully with appropriate semantics on a conceptual level.