Food Process Ontology Requirements

The manuscript presents the main issues that occurs in modelling food processing by means of ontology engineering, with a discussion that focuses on recipe modelling.

Furthermore, the authors offer a survey of the state of the art regarding the topic, showing approaches and solutions adopted by existing ontologies for tackling the problem of food processing and the general problem of modelling planned and unplanned processes with computational ontologies. As the authors say, translating food-related data into machine readable and standardised formats is an ever more important activity in the agri-food domain, which has the potential to contribute to multiple UN's Sustainable Development Goals, e.g., making food chains more productive or reducing food waste (4). The text is also very interesting, comprehensive and a good introduction to the complexity of the subject at issue (1-2). Moreover, it is well written and with good balance of theoretical and example-based presentation (3). As such, the manuscript is certainly worth to be published.

That said, I have minor issues to point out and possible improvements to propose.

* (optional) I would recommend the authors to try to mention concrete use cases to motivate and drive both the topics mentioned in the survey part and the boiled carrots example in the discussion. My feeling is that, especially the non-expert readership would benefit from a clearer connection between what the authors describe and answers to questions like: which queries or competency questions a food processing ontology would be useful for? Which details need to be represented in such an ontology and which others can remain more informal? Which kind of applications would such details support? Examples of questions that might help such a clarifications would be: - an app to search recipes based on (cultural or health-related) dietary needs, with some reasoning capabilities like: finding recipes (not) containing certain ingredients, such as gluten or animal products - list recipes based on their complexity, eg, (not) requiring many steps or complicated operations such as precise quantity proportions or working a pastry dough in the right way.

* Section 1, Introduction. While this is a good opening paragraph, it could be improved by adding citations to support the statements it makes. Example:
Food processing has developed a Janus-faced reputation, offering on the one hand historically- significant transformational changes that make food inexpensive, spoilage-resistant, more nutritious, flavorful, shelf-stable, and convenient [REF]. On the other hand, many of the same features have encouraged overconsumption, unhealthy diets, and food products upsetting human gut microbiome dynamics and health in general [REF]. Academic research into food engineering emerged in the 1950s [REF]... People are increasingly concerned about health issues and environmental impact, and they are also looking for tasty foods, naturalness and practicality [REF]... At the same time, advances in modularized food technologies are enabling food manufacturers to begin to deliver foods according to individual flavor, texture, and other hedonic profiles [REF].

* Figure 1 and 2 are never cited by the text

* Page 4, Left column: (optional) "...with a few additional relations, and doesn’t take an OBO style ontology ecosystem..." you might want to change contractions into more formal extended forms (ie, "does not take").

* Page 10, Left column, typo: "World view: As OBOFoundry had been associated...", space is missing in "OBO Foundry"

* Page 11, typo: a few paragraph number (eg, 3, 3.1.1) are highlighted in yellow in the PDF I've reviewed.

* Page 12: reference to Figure X seems wrong.

* Page 15, L col: typo "In line with its provenance mission, Provo is...", should be PROV-O in place of Provo

* Page 22: (optional) "Not currently included in the recipe model are serving procedures..." The form "serving procedures or derived... are currently not included..." seems more readable

* Page 24, L col: "Given limitations of current OWL reasoner capability,..." does not seem correct or clear. Do you mean something like "Given the limitations of current OWL reasoner capabilities"?

* Page 24, R col: "Having OWL ontologies involved controls the quality..." does not seem a correct sentence and I do not get what it means.

* Page 25, References, typo: numbers in square brackets appear twice.

This manuscript was submitted as 'Survey Article' and should be reviewed along the following dimensions: (1) Suitability as introductory text, targeted at researchers, PhD students, or practitioners, to get started on the covered topic.

As noted in the abstract:
“This work focuses on the ontology components - object and data properties and annotations - needed to model food processes or more general process modelling within the context of the Open Biological and Biomedical Ontology Foundry and congruent ontologies.”

It includes an excellent introduction to the space, then reviews and discusses key ontologies modeling time notions (essential for process modeling) or process related notions from OBO and more general domains (W3C), and pulls out key notions from those ontologies into a coherent and fairly rich ontology for describing processes (primarily planned) involved in food production and also provides enhancements to information notions for describing plans for food production processes (recipes).

(2) How comprehensive and how balanced is the presentation and coverage.

While one could argue that a few important references relevant to process ontologies were not included (e.g. PSL – Process Specification Language, or IDEF0), they arrived at a good balance of comprehensiveness versus document size by focusing on well-known ontology references from either the OBO and Semantic Web communities, which none-the-less included the essential notions for modeling planned processes for food. With respect to balance: I had a real sense that the target audience for this manuscript was primarily the OBO community (which is consistent with the abstract), with perhaps some bias against non-Top Level Ontology based or non-realist based ontologies manifesting in the tone or direction of small parts of the discussion of Prov-o and SOSA. Perhaps I am too sensitive to such things, but in any case, it was a quite small proportion of the paper where I sensed this, and in neither case did this seem to affect decisions about which ontology components to incorporate into their ontology for food processes and recipes.

(3) Readability and clarity of the presentation.

I found this paper to be quite readable. This was due to the text sticking to the main goals of the paper, with descriptions providing just enough detail to describe notions or points (with an exception or two noted in my comments) while not becoming overly verbose. The examples and figures were quite helpful and well explained in the text, with keys explaining their nomenclature. Overall, a very clear presentation of the ideas.

(4) Importance of the covered material to the broader Semantic Web community.

Planned processes are a key element of many human endeavors. Planning, scheduling, and executing such processes in technical areas such as in manufacturing, electric power grids, and in agriculture and food production can impact safety, sustainability, and quality of life. This work brings together elements from a number of ontologies from the OBO and W3C communities to demonstrate a coherent way to model these processes and the plans for them using OWL. While there are many details still to fill in and work out regarding the particulars such as specializations sub-properties of process participants to support planning and automation, it could provide coherent base for further research across semantic web user domains to address those details, and this would be quite important. Furthermore, this paper makes two important additional contributions, 1) proposing a way to fit OWL-Time into the OBO framework such that temporal relations among component processes can be easily expressed (though they didn’t quite finish the job on this in the initial version I reviewed) and 2) pointing out the gap in current observation/measurement ontologies such as SOSA and elements of OBO in providing a simple and standard means of specifying the path from an entity being observed to the quality of the property being measured or observed.

Please also assess the data file provided by the authors under “Long-term stable URL for resources”. In particular, assess (A) whether the data file is well organized and in particular contains a README file which makes it easy for you to assess the data, (B) whether the provided resources appear to be complete for replication of experiments, and if not, why, (C) whether the chosen repository, if it is not GitHub, Figshare or Zenodo, is appropriate for long-term repository discoverability, and (4) whether the provided data artifacts are complete. Please refer to the reviewer instructions and the FAQ for further information.

I found no data file associated with this paper, though I didn't expect one. It was a survey of existing ontologies, and this didn't include any testing or evaluation against sample or reference data sets.

*Detailed comments for authors*

Comments requiring action

1. There is a disconnect between the sentence introducing the bulleted list starting on page 2 and the list itself. It was so jarring that I thought I had somehow jumped a page or missed some paragraphs as I read through the first couple of items on the list. The list seems to be a collection of different kinds of things, most of which are not “aims”. I don’t quite understand how these are all related or how some relate or could relate to the work described in this paper (particularly “The kitchen as a database”). Are these all things that have been enabled or helped by food ontologies in the past or would be enabled or helped by a food process ontology that can address any process elements present in field to fork? A better transition into this list would help, along with a reworking of the descriptions of the first and second items to better fit in the list and explain how each is enabled food process ontology. It’s not clear to me that “the kitchen as a database” fits in the list at all, as it is presently written.

2. There are several important notions in the following sentence from the first full para on page 3:
Ideally, we can model both unplanned processes that we find in the physical world, which themselves might be harnessed by a planned process involving one or more objectives, a plan or detailed protocol to achieve, and which operates on inputs and outputs of material entities or data.
But it is hard to parse the meaning of some of this when it is all in one sentence. Suggest splitting into multiple sentences leading to new text for this sentence to the present colon something like:
Ideally, we can model both unplanned processes that we find in the physical world and the planned processes that harness those unplanned processes to address one or more objectives. Such planned processes execute a plan or detailed protocol to operate on inputs and produce outputs that are material entities or data. A planned process may involve intrinsic or extrinsic transformations and may:

3. The modern version of BPMN is not only capable of illustrating processes. It is also used for creating process models that can then be tested and executed based on a standard semantics. To reflect this, please change “… standard graphical notation for illustrating …” in the description of BPMN on page 4 to “… standard metamodel and graphical notation for modeling and depicting …”.
4. In the process perspective discussion on page 6, “lists of assays” sounds like one or more information entities. How about “sets of assays” instead or simply “assays”?
5. Remove ‘ (aka “process step” ’ from after ‘A planned process’ and before the “:” in the bulleted list on page 8. As currently written, this bullet seems to equate/conflate planned process with component process (syn ‘process step’). I found this very confusing. Not all planned processes fill the role of process step, subprocess, or component process. Please also remove “step” from the text “If a planned process step” in the second sentence of the bullet. If you want to associate the Planned Process notion described here with the Process Step class notion in the DDI process control schema, please qualify the “process step” term in the bullet text with a “ddi:” prefix and put the resulting qualified term at the end of your list of synonyms, and not in the bullet label.
6. The text in the body of page 12 has references to (Figure X). I assume you mean Figure 11. Please fix.
7. The References section at the end of the paper includes two reference identifiers for each reference, e.g. [1] [1]. Maybe one set was autogenerated, and another manually inserted? Please fix.
8. Suggest changing “… a more patchwork set of reference materials near the end of page 9 to “…a patchwork of reference materials…”.
9. In the discussion on weaknesses in temporal properties for planned processes at the end of page 12 and the beginning of page 13, the text says: ‘BFO explains “one dimensional temporal region” as “the temporal region during which a process occurs”’. In BFO 2020, the quoted text is just an example provided as an annotation for ‘one-dimensional temporal region’, so this would be more clearly stated by replacing “BFO explains” with “BFO exemplifies “.
10. More importantly, BFO does include an ‘occupies temporal region’ object property which can be used to connect a process to the temporal region during which the process occurs. Then ‘has first instant’ and ‘has last instant’ would identify the beginning and end times of the interval and therefore the process. So, process time modeling in BFO goes a little further than you describe, but what seems still missing from BFO+RO+COB is any way to express the duration of the time interval.
As I was reading this, I was wondering if you were concerned about the trouble this indirection through temporal interval would cause for reasoning and how it may complicate expressing acceptable ranges for temporal aspects in the plans for processes.
11. In the first bullet near the beginning of section 3.2 OWL-Time, it reads “An ability to express durations during which a process is carried out.” ISO 8601 is the normative reference defining time notions and terms. It defines duration as: “non-negative quantity attributed to a time interval, the value of which is equal to the difference between the time points of the final instant and the initial instant of the time interval, when the time points are quantitative marks”. In other words, durations express the temporal size (magnitude) of a time interval. The quoted text from your bullet doesn’t seem to be using duration in this way. I wasn’t sure whether you were really referring to time interval here or you were actually interested in the temporal magnitude (duration) of the process. If it’s the former, then please replace “duration” with either “time interval” or “time period”. If it’s the latter, then replace “durations during which” with “durations of the time interval during which”.
12. time:TemporalEntity is a unionOf (time:Instant, time:Interval) which appears to be equivalent to bfo:’temporal region’ (obo:BFO_0000008) in BFO2020. But in BFO, bfo:’temporal region’ is disjointWith(process, ‘process boundary’). But the text the section of this paper on OWL-Time seems to treat time:TemporalEntity as equivalent to bfo:occurrent or at least bfo:process. Have the OBO Foundry folks relaxed the monohierarchy rules and removed the disjointness constraints among siblings to allow this? Otherwise, perhaps the reuse of owl-time with BFO based ontologies could benefit from adaptation of owl-time such as decoupling some of its properties for interval and applying them higher up the BFO hierarchy to occurrent, and adding equivalent Allen style object properties for bfo:process to those that are in owl-time for time:ProperInterval. An explicit discussion of this in the paper would be a significant contribution that would be quite useful for BFO user communities, such as IOF. It would also strengthen the comments that mention OWL-Time later.
13. Need a reference for or explanation of SWIRL which is mentioned in the section on OWL-Time.
14. Figure 18 shows time:hasDuration as having a range of TemporalEntity. It actually has a range of Duration, which represents the temporal magnitude of a TimeEntity, and not a TimeEntity itself. The simplest fix would be to simply remove time:hasDuration from the list of properties on the bottom right of the figure.
15. The OWL-Time properties shown in the Data Property section of the Process ontology relation comparison table (Fig.20) are all actually ObjectProperties. Not sure how you want to address this. Maybe change the label for that section of the table to Data Property / Object Property, since the provo properties do appear to be DataProperties.
16. In your description of sosa:Sensor in the bullets describing PO2, you leave out the aspects of the description from SOSA that really distinguishes sosa:Sensor from a more general device that an Actuator is a subClassOf. Suggest you include at least part of the next sentence of the description in SOSA which reads: “Sensors respond to a Stimulus, e.g., a change in the environment, or Input data composed from the Results of prior Observations, and generate a Result.” This could be distilled down to: Sensors respond to some input and generate a Result.
17. The key for Fig. 22: A FoodOn recipe model prototype, change the “other relation” arrow to black to match its use in the figure, rather than blue.
18. There appears to be a couple of small errors in Fig. 23: Partial Boiled Carrots recipe plan specification. The instance ids for the devices in the device set should probably be x_stove1 for the [cooking stove] and x_pot1 for the [container]. This would then match the naming convention used in other figures.
19. I think that, very near the end of page 22, you meant to say that “The result of the process is shown on the right …” rather than “left” as the sentence currently reads.
20. There no references to it, but there doesn’t seem to be a Fig. 25.

Other comments
A. I was curious to see the broad nature of the notion input in this paper. Most of the process modeling formalisms and models that I have seen separate 1) the material or energy transformed or consumed from 2) the actors and equipment that must participate in a process for it to properly execute (be realized). The term Inputs often only refer to 1 or a subset of 1. From later in the paper, subproperties of has participant include: has specified input, has performer, and has specified output. Do you have any finer distinctions (subproperties) among has inputs? Are performers also specified inputs? Is a Computer Numerically Controlled (CNC) machine tool with an operator a performer? What about a completely manual machine tool (it still performs the cutting, but under the manual control of the operator by turning cranks and knobs, etc.)? How do you draw the line between durable process resources (machines, equipment, tools, etc.) that are merely specified inputs and those that are performers? Have people defined roles for this?
B. Another relevant reference for process modeling is IDEF0. While it only supports processes (as Activities), subprocesses and the “flows” between them, it defines roles for Inputs, Outputs, Controls, and Mechanisms that are relevant to my above comment about distinguishing Inputs from Mechanisms, and to the notion of Controls as mentioned on page 5 in the discussion comparing autonomous-oriented and control-oriented process models.
C. I am surprised to see storage and shipping categorized as Transformation processes. What is transformed in a storage process? For me transformation processes change some characteristic, identity, or classification of an entity, so storage or shipping are not necessarily transformation processes (though natural processes may occur during those processes that transform the entity being stored or shipped).
D. On page 8, a bulleted list defines “…our use of terms within the context of a generic process model and within the scope of OWL ontology development…” I understand that to mean you are creating defined terms for the “generic process model” that you will use in the rest of the paper, when not discussing terms from some other source.
E. Interesting that in the 2.1.4. Process Model nomenclature section (on page 9) you have chosen measure (aka measurement) and observation to denote the information element resulting from an observation process. The use of measurement and observation for these entities is common in my domains, but I had noted that SOSA and OBIE(sp?) had gone the other way, using these terms for the processes and calling the information entities produced by such processes results. I was planning to change the way we did this in our models to match these ontologies.
F. I think the key point about PROV-O’s applicability to describing process is made in your sentence reading: “In line with its provenance mission, Provo is backward looking…” The purpose of provo is to document the history of something where the things of interest were usually information objects. It’s tailored to that purpose and to its motivating use cases, and as a result it is not a good fit for general process description or IMO for provenance/pedigree outside its target domain and uses. Not sure what the value is of the OBO-based toy model of publishing though. Is it to show how the elements of your process model could apply to the same sorts of processes that Provo is intended to document?
G. I found the text in the first discussion sosa:FeatureOfInterest that reads, “SOSA’s flexible sosa:FeatureOfInterest sidesteps the direct naming of the thing whose feature is being sensed (the domain-specific entity and its qualities), leaving it up to 3rd party ontologies to do so.” to be a bit a bit flip (though, I have no ownership in SOSA/SSN, let alone of this element of it). sosa:FeatureOfInterest is an abstraction of the thing being observed role, though its name (derived from OGC specs) and its abstract/vague description in SOSA/SSN can lead to confusion about it. The intent with an ontology like this is to minimize commitment so it can be composed with a variety of other ontologies, so yes, users of the ontology have to fit the entities being observed and the properties of those entities that they want to measure in their domain and use cases to FeatureOfInterest and ObservableProperty respectively.
Your later point though, about the location of the quality “left to the implementers to structure”, I think is quite important. All of the observation or measurement models/ontologies I have reviewed are lacking in providing the infrastructure to describe a full path from the EntityBeingObserved to the sosa:ObservableProperty. These paths can be quite complex for product specifications in manufacturing, but even simpler use cases we have seen for agricultural sampling are awkward to support with only the two components available in the models we have reviewed.
H. Is there any special meaning to the bolded text in Fig. 23 on page 22?
I. On page 24, you discuss using a tracking identifier “… to denote a given material entity anywhere within a process input / output matrix (see publishing example Figure 16), effectively marking a set of instance entities as roughly the same continuant while avoiding OWL equivalency logic directly.” Indeed. But Fig.16 is probably not the best example for this; one of the figures from the boiling carrots example would be far simpler fit to your description above. The entities denoted by ‘book identifier’ in Fig. 16 are shown as Material Entities. Presumably then, ‘draft book’ is a written/printed copy, and ‘printed book’ is a physical printed book. The entities for a book example that roughly correspond to each other might rather be the ‘information content entities’ associated with these MaterialEntities, and they could represent “the work” at different phases of its lifecycle. But I don’t think you want to go there.

The paper presents an analysis of some selected ontologies: some of them not born for the specific food domain (e.g., SSN/SOSA, OWL-TIme ontology), others with different objectives with respect to the needs of the food domain (e.g., PROV-O), and some more targeting the domain of analysis (e.g., OBO foundry ontologies). These ontologies are analysed with the goal of identifying elements that can be relevant for harmonising food processes modelling and in general processes modelling within OBO foundry ontologies. The authors conclude the analysis with a discussion, specifically related to food recipes, seen as food processes; a prototype with the FoodOn ontology is presented.

The paper has been classified as a survey paper that, according to the rules and guidelines of Semantic Web Journal are papers that should have the potential to become well-known, highest quality introductory and overview texts, and should be suitable as introductory text, targeting researchers, PhD students to get started on the covered topic.
From my perspective, the paper in its current form definitely lacks these elements and my overall judgement is reject.

I explain below the reasons for this evaluation by providing a first overall feedback on the paper and then detailed comments on specific sentences of the paper, that further justify my overall judgement.

In general the paper is in scope of the special issue and presents an interesting topic. However, a number of issues are highlighted in the following.

The paper seems a draft: number of sections highlighted in yellow, different occurrences of “figure X”, as a placeholder for including a figure reference, links to some web sites that seem once again placeholders for a successive inclusion in the bibliography. Many figures are included in the paper but they are never referenced in the text and so never explained.
There are sometimes bibliographic elements in red ([1]). There is the indication of some sections that are not part of the paper (e.g., “OWL reasoning in a process ontology”). The bibliography includes repeated numbering, probably due to some issues on exporting in PDF from a Microsoft Word document.

In addition, I have concerns regarding:
1) the survey analysis;
2) the overall narrative of the paper
3) some semantic modelling proposed in the paper

1) In particular, I believe the analysis, which should be the core of the paper, is not sufficiently clear and this limits the possibility for this paper to be “suitable as introductory text, targeting researchers, PhD students to get started on the covered topic.”, as required by the journal guidelines.
Section 2 starts the assessment of the state of the art but there is no a clear description of the methodology followed for the assessment. This is a crucial point: from my perspective, a survey paper should clearly explain the method that has been chosen for analysing specific works, with a dedicated section that for instance explains the different approaches to process modelling and the criteria of the analysis. The paper then should follow those criteria in the description of the state of the art, and in the comparative analysis. In addition, the paper is entitled “Food process ontology requirements” but these requirements, that probably are important to introduce the different criteria of the analysis, are briefly introduced as last section of the paper. This is quite weird.
In the conclusions, the authors summarise the work presented as “a gap analysis”, but to be honest I do not see a truly gap analysis, I just see some general descriptions of processes with some references to the OBO ontologies, and some ontologies such as OWL-Time that is definitely important but, I wouldn’t have categorised it as the core for the food process modelling. Once again, the lack of a methodology for the assessment prevents readers to fully understand the reasoning followed by the authors.
In addition, I believe that some important ontologies are missing in the analysis. For instance Dolce-Ultra light, that, at some point, is mentioned as only a prefix of a few classes aligned with SOSA/SSN, provides interesting elements for modelling plan, workflow etc. but this is not considered in the paper.
In the ontology world related to food, this work published at ISWC - https://link.springer.com/chapter/10.1007%2F978-3-319-46547-0_18 appears just searching in google. This is not part of the bibliography. It seems including some classes for modelling characteristics that are cited in the conclusions of the paper as future work.
Agrovoc is another important thesaurus of the sector that is missing in the paper. Is there a reason?
Also, ontology design patterns, related to process/activity, are never considered in the paper but I believe they can be useful in specific cases of the discussion.
Finally, the discussion section, that I supposed it was related to the state of the art analysis, is instead a proposal of a model for the specific domain of the food recipes. In this section, a prototype in FoodOn is presented. FoodOn is never referenced in the paper but there is a publication of this work available https://doi.org/10.1038/s41538-018-0032-6 . It is a work of some of the authors and surprisingly is not part of the bibliography of this work.

2) The narrative of the paper should be also revised for me. The description is not organic and linear. The paper is full of paragraphs that suddenly introduce things that are never referenced in the bibliography, and for which, in some cases, it is difficult to understand why introducing them. For instance, in section 3.6 the authors suddenly speak of food ontologies after a long discussion on PROV-O and sensor-based ontologies like SSN/SOSA; again, the scope of the discussion on publishing book, in the form used to present it, is unclear. This further confirms that, from my point of view, the overall approach to the survey should be revised.
In addition, there are sentences that are never motivated. For instance: “other ontologies or model framework were considered but not given in-depth review”. That’s might be ok, but why? Only for FTTO - Food Ontology for Traceability Purpose - the authors motivate the exclusion from the analysis because the ontology does not take the same approach as the one used in OBO. BTW: I am not sure this is a valid reason to exclude it in a survey paper regarding food processes, if you considered PROV-O for instance in the analysis. In addition, mixing ontologies and notations such as BPMN sound strange, unless very well motivated reasons in a methodology for the analysis are introduced.

3) Finally, there are some figures that probably propose some modelling. In one of them I have some doubts, from a semantic perspective, about the application of OWL-Time properties. In particular, in OWL Time ontology time:hasBeginning and time:hasEnd have as domain time:TermporalEntity which is the union of time:Instant or time:Interval.
In Figure 10, these properties are linked to a process. So the authors are saying that a process is a time:TemporalEntity (union of time:Instant, time:TimeInterval) and I do not think is correct. Same applies for temporalDuration. From my perspective a process has some temporal entity not is a temporal entity in the sense defined in OWL time ontology. In fact you say afterwards a process can have a temporal duration. With this respect, referring to a foundational ontology such as DOLCE-ultra light can be beneficial here to clarify these concepts.

Some other detailed comments:

Introduction section: “‪but in this‬ ‪paper we focus on the more general language required‬ ‪for modelling.” —‬> what do you mean precisely? It should be better explained.

Fig. 1 is never referenced in the text. It seems referring to the discussion of the process objectives; it is not clear if it is something derived from state of the art or an elaboration of the authors.

Fig. 2 never referenced in the paper

CyVerse and Galaxy are not included in the bibliography

Why ‪Petri nets are mentioned in a section that is the state of the art on Ontology-based processing models? For me it is confusing.‬
‬
Blockchain suddenly ‬appear in the paper with no reference

“‪Examples are detailed in model applications below‬” —> better say “in Figure 4” rather than “below”

“‪It‬ ‪must be completed before proceeding to the next step.‬ ‪In this respect, a control-oriented process of “steps”‬ ‪forces autonomous processes into a linear sequence.‬” —> it seems to me that this sentence is a very simplified version of process models. This linear sequence does not take into account other more complex processes where parallelism is also possible. I did not see such a discussion in section 2.1.3 if the objective here is to describe process models.

PROCO is not referenced

OBO main references should be in the bibliography

SSSOM should be in the bibliography

All the ontologies mentioned in section 3.1 should be in the bibliography

“It‪ may seem confusing to explore an ontology entity‬ ‪and see another ontology’s terms used in one or a few‬ ‪of its axioms”‬ —> why confusing?

Which is ‪(Figure X)?? Figure 11?‬

Fig.12 never referenced in the text.

“‪Exact 2 approaches‬ t‪his with an intermediate level of process model‬ ‪containing input and output relations and materials,‬ ‪and contains over 90 “experimental actions” one finds‬ ‪in a laboratory context, and a general “has proposition”‬ ‪object relation that connects an action to various‬ ‪contextual objects such as expressing a condition for‬ ‪an action to proceed, a duration, or a “protocol method”‬ ‪to follow.‬” —> this sentence is difficult to read. I would suggest it to re-write it.

In some parts, the discussion about PROV-O sounds to me not really appropriate. For instance: “although it awkwardly refrains from providing features that support a general process model mandate.” —> ontologies are created to meet specific requirements in domain modelling. Don't you think that probably the objectives of the model were not exactly what you have in mind in this document?
In addition, “PROV-O introduces its own time data properties rather than reusing OWL-Time” —> said like that, it sounds like a precise modelling choice of PROV-O. Actually, there is a historical reason for this and it should probably be mentioned (for long time OWL-Time was a W3C note only, and its adoption in ontologies like PROV-O has been compromised by the uncertain status of the OWL-Time specifications. Now OWL-Time is standardised but this happened after the publication as Web standard of PROV-O).

“‪OBI in contrast has a planned process‬ ‪execute a plan specification or protocol.” —> ‬do you mean “O‪BI in contrast has a planned process‬ that ‪executes a plan specification or protocol.” If so, I would suggest you to re-write it in this way.‬
‬
“modelling th process of‬ ‪observations‬” —> the*

Fig. 17 never referenced in the text.

“‪see 5.3. Events‬ ‪versus records in‬ ‪https://doi.org/10.1016/j.websem.2018.06.003)‬ “ —> I would say “the interested readers can refer to [X] for a discussion on events versus records modelling for the observations” and [X] should be in the bibliography.

“‪assay ‘has‬ ‪output’ some ‘measurement datum’ and ‘is about’‬ ‪some entity”‬ —> Well, here you put not only the about what, but also the output, In SSN/SOSA the output is the result linked to the observation, that in turn is linked to FeatureOfInterest.

“‪The different domain sub-ontologies that specialize the PO2 core model will be‬ ‪developed in different projects (for example, dairy,‬ ‪meat, and bakery food production) to represent the‬ ‪characteristics of foods during their manufacturing‬ ‪process.‬” —> before you said that the PO2 ontology has two layers: core and domain. Is the domain layer available? According to this sentence it seems no. This should be clarified.

In Figure 19 it would have been beneficial including prefixes to understand which classes are from OBO network of ontologies and specifically PO2 and which ones from SOSA/SSN.

Fig. 20 seems a quite important figure that compares the analysed ontologies, and it is never explained in the text.

I really do not get the point, with respect to the overall discussion, on the impact on SPARQL queries for having general properties In some context general properties can be effective and used and I do not see such a complex SPARQL query.

“‪We turn to applying an ontology-friendly modeling‬ ‪paradigm to recipes which are generally a set of‬ …” —> what is an ontology-friendly modelling paradigm? What do you mean?

“‪Its fundamental meaning is the putting together‬ ‪of components in appropriate relationships or‬ ‪structures, according to a formula‬ ‪(https://en.wikipedia.org/wiki/Formulation)‬” —> reference in the text to wikipedia on a generic term? I would avoid this.

“‪Figure X illustrates ..” ‬ —> are you referring to Fig. 26?

‪ENVO‬ —> what’s that? The Environment Ontology? Explain the acronym and put a reference in the bibliography