A Behavioristic Semantic Approach to Blockchain-based E-Commerce

This paper describes the design and evaluation of three ontologies for describing decentralised e-commerce transactions, in particular those supported by Blockhains (Ethereum).

I think the paper needs to greatly improve writing and presentation, and clarify what is its methodological approach. Detailed comments per section follow.

SECTION 1:

Fells short in motivating why these ontologies are needed. S1.1 cites "simplification and automation for the tasks of probing the blockchain for desired activity or token related features", and "Smart Contract exchanging specific types of tokens and trading conditions" but I don't see in turn what those important, e.g., Blockchains can already be probed for token related features (not Semantically, mind you, but yes in general through their APIs).

The paper mentions several goals throughout that are not always consistent between them or with, in S1.1 we see "lay out the foundations for a clear and unambiguous semantic representation of the blockchain in the area of e-commerce", later on we read "the conjunction of business and blockhain systems from an epistemological point of view, hence the goal for the present work". That one is quite broad and I am not sure how the concrete contribution contributes to it. Furthermore, there is an unexplained jump from e-commerce to "business" in general.

There is a full paragraph on NFTs, but it is not clear at all what is the connection of that topic with the remainder of the text, that talks about blockchains in general. One would think that in an e-commerce context, payment with cryptocurrency is what is most needed, but again, as up to this point there is not enough clarity on the motivations, we are not certain where the development of the ontologies are headed.

In section 1.3 the contribution is stated as "The present article reports on the foundational results of POC4COMMERCE". There is not sufficient explanation of how you get from POC4COMMERCE and ONTOCHAIN to the goals in 1.1.

From section 1.2 I also highglight the use of the "behaviouristic approach" [10]. As this in a section named "Chosen approaches" I take it as a hint on the methodology followed to develop the proposed extensions. I'll come back to this later because I did not see very well how this was applied later.

SECTION 2:

If the goal of the paper pertains to blockchain and e-commerce, I don't see why DARPA and OWL-S should be mentioned, or ontologies about IoT services. In terms of the section of automated agents, do you expect automated agents to be doing the blockchain-based e-commerce? None of the goals in section 1 seem to go into this direction. A possible improvement could be the introduction of an explicit sub-section to group what are the works mentioned as "things that have successfully used the same methodology we are using" to separate them from "other ontologies on each of the levels we are operating".

A few vocabularies for product descriptions are mentioned, but for commerce, only GoodRelations is covered. If the goal is especifically e-commerce, I'm missing works specific to it, like https://link.springer.com/chapter/10.1007/978-3-540-36562-4_51 or https://link.springer.com/chapter/10.1007/978-3-319-46547-0_27, and even schema.org, which is the one used by commercial search engines for this purpose.

In terms of Blockchains and Semantic Web, or Blockchain and ontologies, I'm missing https://link.springer.com/chapter/10.1007/978-3-030-36691-9_19 and https://ieeexplore.ieee.org/abstract/document/8783113

SECTION 3:

Comprehensive, see it as an extension of the related work.

SECTION 4:

This is where I have a methodological issue, commercial actors, offers, products and tokens emited as witnesses of exchanged assets are mentioned as "essentials". How this was established? If the behaviouristic approach from [10] is followed where are the goals and dependencies between each of the actors?

OC-Found is said to "model the stakeholders of the blockchain-oriented commerce ecosystem and exploiting the OASIS ontology", but then in section 4.1 we found that what it does is to extend OASIS with a semantic representation of "digital identities, supply chains and quality valuation mechanisms", these are not stakeholders. As we don't know the goals and dependencies of these actors, we cannot establish why you need to have digital identities (first time this is mentioned, unclear if this relates to decentralised Ids or they could also be centralised certificates). If this is just about e-commerce, why do I need a supply chain? For me, e-commerce is limited to a product in a warehouse that is bought by a customer. It looks like here the scope of that is broader, this might be clear in the POC4COMMERCE project, but it is not in this paper.

There is a quite comprehensive description of the OASIS templates, but if I understand correctly, they are not a contribution of this paper. Moving from the UML diagrams that explain the high level templates to a Protege screenshot for the actual contributions is not very clear. Similar to the comment above, what are the PlanExecution, GoalExecution etc. _instances_ that support the inclusion of the classes and properties in Figure 4 and Figure 5? If this is absent, then the oasis:Behaviour and oasis:Agent propertier are merely re-used in OC-Found, not being used as a behaviouristic method to develop OC-Found. I also wonder if ontologies for supply chains should have been covered in the Related Work section in the same detail that, eg. https://doi.org/10.1016/j.dss.2013.09.007

SupplyChainProofOfWorkActivity. Not all digital tokens and blockchains use proof of work. Is this is related to transaction confirmation as with a centralised payment provider, it would probably need to be called like that.

Not clear why "supply changes have to be considered as interchangeable supply chains"

Overall, I have similar comments for the OC-Commerce, it is unclear what method was used to decide the additional classes (bargaining activities) to be added. Here we observe a bit more clarity in terms of the agent behaviours (p.18) and the Figures provide better context. It seems that the example may be used to illustrate the missing requirements

For Section 4.3, not clear how an extension of Blondie "provides a behaviouristic vision of commercial activities". Again, some agent behaviours and goals are missing to understand the need for Smart Contracts. One can infer that a seller or producer agent may create NFTs or digital certificates of products that are then exchanged on an e-commerce context, but it is not clear how the other types of tokens could be used here, especially in connection with the other parts of the ontology stack. Are they to be used as payment?

One question is what type of UML diagrams are these, they look like Class Diagrams, but in that case shouldn't agents be actors?

SECTION 5:

The quantitative metrics are fine. As with other parts of the paper, the main question is how the competency questions were derived. Usually, these are developed together with the other functional requirements, not after the development of the ontology.

SECTION 6:

I don't think a "map" is the right word here, I think "model" is best. I also found very confusing some usages of the word "exploiting".

I appreciate the existence of such a section, the only issue I find is that app orders include an "address for a Snart Contract associated with the application." Following from the outline of what an application is , I don't see why a computer program to be executed in the Cloud should have a Smart Contract and for what. It is also unclear why worker pool orders are modeled as supply chains.

SECTION 7:

The conclusion focuses mostly on the blockchain aspects, it says that "this article achieves a formal semantic representation capturing the smart contracts on the blockchain as well as the activities carried out on it". But OC-Ethereum is restricted to Ethereum and only one third of the overall contribution. I also disagree that extending an ontology "facilitates the understanding of blockchain contexts". There is no mention about any of the goals in section 1 and if they were achieved. As the efforts are focused on Ethereum and that both OC-Commerce and OC-Found model things that are e-commerce related and can perfectly exist without blockchain, I don't think this can be sold as supporting blockchain-based e-commerce.

In terms of resources, the repository includes the ontologies, but it does not include any of the tests (competency questions). I was also expecting the model demonstration of the iExec example to be available.

Minor edits:

S1.3: founded the ONTOCHAIN -> funded the ONTOCHAIN

There are several places where you use the word "delivering" where I think you mean "delivery"
Fig 7: DigitalIndentity -> DigitalIdentity

In this work, the authors extend the OASIS (agents), GoodRelations (commerce) and BLONDiE (blockchain) ontologies with the purpose of making e-commerce more accessible. The work is accurate and well written.

The authors make a good evaluation of each ontology (using different reasoners and defining and resolving some competency questions in each ontology), although more work is needed and is indicated in the conclusions section. However, I have the following comments:

- In the introductions, the authors make the following statement: "One of the most popular applications of Turing-complete blockchains such as Ethereum to business activities is the smart contract." So my question is: Are there any applications of Turing-complete blockchains beyond Smart Contracts?

-The types of contracts could be better detailed. The ERC20, ERC721 and ERC1155 protocols are mentioned without explaining, for example, what the semi-fungible tokens are.

-Will each ontology be published in a permanent identifier such as w3id (https://w3id.org/) in order to reuse it?

And some typos:

-In Figure 1, Thing should be owl:Thing.

-On page 1, "continuos" -> "continuous"

-On page 42 -> "... to a dataset orde," -> "... to a dataset order,"

-Just a comment. In the 3.3 section, the authors mention Bitcoin, Ethereum and Hyperledger and their main competencies, such as: "Who mined a specific block" and "how many total coins were transferred on a block". Hyperledger has no mining blocks or coins. It is a simple comment, and it doesn't matter since the work is focused on Ethereum.

Throughout their paper, the authors present ONTOCHAIN, a novel ontology architecture to represent blockchain-oriented e-commerce, from its supply chain and identification of traded objects (with the OC-Found ontology) to the offerings and the behaviour of the agents that are responsible for their management (with the OC-Commerce ontology). The OC-Ethereum ontology links the commercial assets to tokens, their characteristics and trading. ONTOCHAIN resorts to existing known ontologies: OC-Found integrates the OASIS ontology to describe agents by their behaviour; OC-Commerce extends the GoodRelations OWL vocabulary for products and services; finally, OC-Ethereum includes concepts from the BLONDiE ontology. After a detailed examination of the concept models the ONTOCHAIN architecture consists of, the authors evaluate their framework by means of measures that quantitatively assess the quality of their structures. On top of that, they show SPARQL queries to extract meaningful information from their ontologies. Finally, they propose a case study in which a marketplace for cloud resources named iExec.

The paper is written very well, aside from a few typos (an excerpt listed below). The authors remarkably dedicate quite some effort to the detailed explanation of the diverse conceptual components that contribute to the ONTOCHAIN framework. The integration work with established ontologies in adjacent semantic areas is noticeable. Indeed, the Relationship Richness highlights this quality. The approach the authors followed to evaluate and demonstrate their work stands out for its systematicity. All in all, it is quite a good article to read and proposes many interesting reflections and design choices.

On the downside, it seems that the authors have overlooked the behavioural aspect of agents and smart contracts to depict the expected interactions. Several research papers have been published on this topic from the perspective of work and business processes, which represent interaction and collaboration dynamics among parties to deliver a business and work outcome. The authors can consider as a starting point the following contributions in the field:
https://link.springer.com/article/10.1007/s12599-009-0078-8
https://link.springer.com/chapter/10.1007/978-3-540-30075-5_2
https://aisel.aisnet.org/wi2017/track03/paper/6/
As the full specification of entire interactions between parties may be either overly specific (scarcely generalisable) or too general (thus barely representative of the interactions), the authors can also consider declarative approaches to process modelling as in the following papers.
https://link.springer.com/chapter/10.1007/978-3-540-76888-3_83
https://link.springer.com/chapter/10.1007/978-3-030-21297-1_6
Last but not least, ontologies for process execution data extraction have also been proposed, which could be meaningful for their purpose and intent of documenting the full-stack e-commerce processes, from the supply chain down to the purchase orders, for traceability and auditability purposes.
https://link.springer.com/chapter/10.1007/978-3-319-42887-1_12
https://link.springer.com/chapter/10.1007/978-3-642-33203-6_4 

In that regard, it is also worth mentioning that the notion of supply chain appears to be not fully integrated with the core contribution of the paper. The authors should clarify better why they extend the span of the ontology that far, considering that it is presented as centred around e-commerce. In this way, it looks like focussing on the e-commerce supply chain (where the actual commercial phase is at the end of the process architecture).

Blockchain for supply chains is especially relevant for traceability purposes (see, e.g., https://doi.org/10.1016/j.cie.2020.106895 for a survey and, more related to processes and thus behavioural aspects, https://doi.org/10.1007/978-3-319-94214-8_4). However, tracking the origin and byproducts that lead to the final outcome appears not among the core queries. Notice that the query should build a concatenation of events in case, which depicts the course of actions leading to the end of the supply chain e-commerce process. The authors should elaborate more on this aspect. They are also recommended not to suggest it as future work as it does not seem viable here due to the centrality of this point in the overall context.

Section 1.3 (contributions) read quite impersonal compared to the other sections, to the extent that it remains unclear whether ONTOCHAIN (never mentioned before in the paper) is the ontology the authors propose or is part of the related work. This doubt remains till Section 1.4. Section 1.3 reads a bit like the synopsis of a deliverable for a research project: the authors should rephrase it to make it more suitable for an article.

As the reader can understand from the Acknowledgements section, the authors are directly involved in the European research project that is named the proposed ontology. Interestingly, it involves companies in agri-food e-commerce. Proposing a case study that deals with physical or tangible products could have been appropriate. The presented iExec case, wherein buyers and sellers deal with resources, applications and datasets for cloud computing, reads weakly related to a full-fledged supply chain. As the authors themselves state, “ defining supply chains for these assets would not be appropriate. Consequently, classes representing the app and dataset offerings […] are not
associated with supply chains. Instead, a supply chain is specified for the class representing the worker pool offering.” I would say that claiming workers as part of the purchasable objects in a supply chain leaves quite an uncomfortable feeling in the reader. Ethically, it reads highly questionable as a comparison indeed. The authors are required to add another use case where the role of a proper supply chain is more prominent and avoid the possible misunderstandings that the current one can generate.

The authors should rename the “SupplyChainProofOfWorkActivity” class, as (we read) it “describes the activities related with the process of releasing some proofs of work, such as digital tokens emitted to witness the transferring of ownership of the resource.” As the ontology is intertwined with Ethereum and Bitcoin blockchains, the name clash can disguise the actual sense: Proof of Work is an algorithm behind the rights to publish blocks at a lower abstraction layer.

As new technologies are emerging, the authors should also discuss the extendibility to other blockchains than Bitcoin and Ethereum or Hyperledger Fabric. An ontology should be forward-compatible to limit the risk of deep refactoring with time.

Also, the authors should rephrase the following statement:
“One of the most popular applications of Turing-complete blockchains such as Ethereum”
Ethereum is not a Turing-complete blockchain. Ethereum is a blockchain equipped with an execution environment for programs that are Turing-complete.

A list of typos concludes this review. The authors are requested to proofread their manuscript to fix the ones below and other grammar issues that occur in the document (which is overall well written, yet its considerable length makes typos unavoidably probable).
“Much more recently” → “More recently”
“In consequence” → “As a consequence”
“Related Works” → “Related Work”
“The ontology OASIS” → “The OASIS ontology” (a similar action is required for the following titles)
“OC-FOUND” → “OC-Found” (page 9)
“comprises three” → “consists of three” / “is comprised of three”
“in their turn” → “in turn”
“as first step” → “as a first step”