Transforming SHACL Shape Graphs into HTML Applications for Populating Knowledge Graphs
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsThis is an interesting study. I recommend the following point to help improve the technical quality of the paper.
Define terms and abbreviations at their first use and mention.
The methodology is rigor, but the subprocesses of each process (milestone) of the BPMN model defined in Figure 2 should also be clearly represented, expanded and described.
An algorithmic representation of the proposed method taking into account the systems architecture (Figure 1) will help add more clarity and novelty to the study.
The discussion of the theoretical and practical implications of the study in Section 6 (Discussion) should be done in connection with the existing works of literature.
Author Response
Dear reviewer,
Thank you for your positive feedback and valuable comments. We have carefully addressed those points in the revised manuscript.
Comments1: Define terms and abbreviations at their first use and mention.
Response1:
We corrected the manuscript to define abbreviations at their first use, where we noted it. Additionally, we have a list of abbreviations along with their meaning before the section of references.
Comments2:
The methodology is rigor, but the subprocesses of each process (milestone) of the BPMN model defined in Figure 2 should also be clearly represented, expanded and described.
Response2:
We clarified that the sub-processes mentioned in Figure 2 are further expanded in the next sections, namely, the first sub-process "Insert general restaurant data" is further decomposed in subsection 3.4 , as shown in Figure 4, and the sub-process "Insert dishes in menus" is further decomposed in subsection 3.7, as represented in Figure 5.
Additionally, it is explained that the remaining sub-processes, "Insert restaurant menus" and "Insert dish ingredients", are structurally similar to the "Insert general restaurant data" sub-process and are therefore not discussed in detail in the description.
Comments3:
An algorithmic representation of the proposed method taking into account the systems architecture (Figure 1) will help add more clarity and novelty to the study.
Response3:
A new algorithm is introduced in the section 3. ( System Architecture and Form Application Model Definition) that describes the system architecture. The process is described through two main procedures. The first one represents the design of models, and the second one shows the population of knowledge graphs through the use of a generated application.
Comments4:
The discussion of the theoretical and practical implications of the study in Section 6 (Discussion) should be done in connection with the existing works of literature.
Response4:
We changed the Discussion section to address the tools that we reviewed in the Related Work, and explained where our solution stands in comparison to the previous tools. The practical implications are discussed in line with the defined requirements. It is clarified particularly what research gap we addressed.
Best regards,
authors
Reviewer 2 Report
Comments and Suggestions for AuthorsThis paper presented an approach to generate various HTML-based application codes from the given SHACL shape descriptions. So this work is a "practical" approach to help build up applications using SHACL and ontology-based descriptions.
Strengths:
One of the advantages of this proposed approach is it allows the users to (re)use existing or newly prepared business process models upon BPMN ontology for its automated code generation including the applications logics by the defined business processes in there.
Weakness:
Although this paper did not give solid numerical evaluation results, this paper elaborated how the actual business processes can be written and transformed into application codes using a case scenario example. It would be very nice when the authors gave some reasons not to prepare numerical results in this context as well as mentioning about obtaining experts comments for the quality of the produced codes.
About References:
It would be very nice to have a discussion regarding some recent results regarding the issue of complex model and code generation using Large Language Models. For example, in [A], it has reported that an LLM can be utilized to produce a draft of an ontology. Does this proposed approach combine some LLM-based model and code generation techniques? Which kind of benefits would be have comparing to only using LLM for the code generation?
[A]: Anna Sofia Lippolis, Mohammad Javad Saeedizade, Robin Keskisärkkä, Sara Zuppiroli, Miguel Ceriani, Aldo Gangemi, Eva Blomqvist, and Andrea Giovanni Nuzzolese. 2025. Ontology Generation Using Large Language Models. In The Semantic Web: 22nd European Semantic Web Conference, ESWC 2025, Portoroz, Slovenia, June 1–5, 2025, Proceedings, Part I. Springer-Verlag, Berlin, Heidelberg, 321–341. https://doi.org/10.1007/978-3-031-94575-5_18
About possible improvements of Figures/Tables:
On Figure 3, since it did not include any outer frames in the picture and the form used very flat design for the UI components, we did not notice which area would be included as the form that the authors would mention. Please put some borders for the part of screenshot or put some extra explanations for the components used in the form.
Author Response
Dear reviewer,
Thank you for your positive feedback and valuable comments. We have carefully addressed those points in the revised manuscript as follows.
Strengths:
One of the advantages of this proposed approach is it allows the users to (re)use existing or newly prepared business process models upon BPMN ontology for its automated code generation including the applications logics by the defined business processes in there.
Comments1:
Weakness:
Although this paper did not give solid numerical evaluation results, this paper elaborated how the actual business processes can be written and transformed into application codes using a case scenario example. It would be very nice when the authors gave some reasons not to prepare numerical results in this context as well as mentioning about obtaining experts comments for the quality of the produced codes.
Response1:
The evaluation and automatic generation of models is planned to be part of a subsequent paper. To make it clearer, we changed the discussion part of the paper to clarify that the empirical evaluation with numerical results will be carried out in several categories and is planned for a subsequent paper. We are still in the process of generating test data for the evaluation part.
Comments2:
About References:
It would be very nice to have a discussion regarding some recent results regarding the issue of complex model and code generation using Large Language Models. For example, in [A], it has reported that an LLM can be utilized to produce a draft of an ontology. Does this proposed approach combine some LLM-based model and code generation techniques? Which kind of benefits would be have comparing to only using LLM for the code generation?
[A]: Anna Sofia Lippolis, Mohammad Javad Saeedizade, Robin Keskisärkkä, Sara Zuppiroli, Miguel Ceriani, Aldo Gangemi, Eva Blomqvist, and Andrea Giovanni Nuzzolese. 2025. Ontology Generation Using Large Language Models. In The Semantic Web: 22nd European Semantic Web Conference, ESWC 2025, Portoroz, Slovenia, June 1–5, 2025, Proceedings, Part I. Springer-Verlag, Berlin, Heidelberg, 321–341. https://doi.org/10.1007/978-3-031-94575-5_18
Response2:
Thank you very much for pointing us to this important source. Similar to the generation of ontologies with LLMs in the mentioned paper, we plan to use LLMs for the generation of our application models. We plan to consider automatic generation of interface models (extension of SHACL files) for the next paper. The definition of competency questions as done in the paper you mentioned can is suitable for definition of competency questions for the user interface that will be made by LLM assisted generation of our models.
Benefits of using our models instead of code generation are better structuring and the possibility to check the consistency of the model by using ontology reasoning, which is not possible with just code generation using LLMs. Therefore, this is a topic for a subsequent paper.
Comments3: About possible improvements of Figures/Tables:
On Figure 3, since it did not include any outer frames in the picture and the form used very flat design for the UI components, we did not notice which area would be included as the form that the authors would mention. Please put some borders for the part of screenshot or put some extra explanations for the components used in the form.
Response 3:
We have included a border around the part of the generated form in Figure 3, and explained additionally what UI components are included in the generated HTML form.
Best regards,
authors
Reviewer 3 Report
Comments and Suggestions for AuthorsThe paper presents innovative approaches to generating web applications from SHACL shapes using ontologies and business process modeling, but it also exhibits several weaknesses:
-
Limited Evaluation and Empirical Results: The paper primarily describes the methodology and conceptual framework without providing comprehensive empirical validation, real-world case studies, or performance metrics to demonstrate effectiveness and scalability. This limits confidence in practical applicability.
-
Although the paper mentions an implementation repository on GitHub, it provides limited details about actual system deployment, user interfaces, or integration challenges, which are important for practical adoption.
- Accessive self-citations: The mansuscript contains around 10 self-citations. These are references where the authors cite their own prior work. This much number of self-citation is on a high side.
Author Response
Dear reviewer,
Thank you for your positive feedback and valuable comments. We have carefully addressed those points in the revised manuscript.
Comments1:
The paper presents innovative approaches to generating web applications from SHACL shapes using ontologies and business process modeling, but it also exhibits several weaknesses:
- Limited Evaluation and Empirical Results: The paper primarily describes the methodology and conceptual framework without providing comprehensive empirical validation, real-world case studies, or performance metrics to demonstrate effectiveness and scalability. This limits confidence in practical applicability.
Response1:
The evaluation and automatic generation of models is planned to be part of a subsequent paper. To make it clear, we changed the discussion of the paper to clarify that the empirical evaluation with numerical results will be carried out in several categories and is planned for a subsequent paper. We are still in the process of generating test data for the evaluation part.
Comments2:
- Although the paper mentions an implementation repository on GitHub, it provides limited details about actual system deployment, user interfaces, or integration challenges, which are important for practical adoption.
Response2:
Thank you for your remark. We have modified the application and inserted the explanation as to how to install and start the prototype using the configured Docker containers and without the use of containers. This can be read in the readme file. A more detailed explanation of the architecture of the whole system is added as a brief user guide to the repository (https://github.com/ontosoft/onto-app-models/blob/main/documentation/OntoUI-documentation.pdf)
https://github.com/ontosoft/onto-app-models
Comments 3:
- Accessive self-citations: The mansuscript contains around 10 self-citations. These are references where the authors cite their own prior work. This much number of self-citation is on a high side.
Response3:
It was not our goal to make self-citations, because we referenced two previous papers which introduced and explained the OBOP ontology used in this paper, and the third reference is a paper of prof. Paulheim with the user interface ontology. For instance, to avoid self-citations, we even excluded the following source, where an application of our approach in one setting was discussed:
Arising Internet of Everything: Business Modeling and Architecture for Smart Cities, Chapter 7 in Recent Developments in Engineering Research Vol. 8
We think that you might have an impression of invasive self-citations because links to repositories where the OBOP ontology, BBO ontology, application prototype and other similar sources are stored are also included in references. We have indeed put those links in citations instead of footnotes. (examplse Rutesic, P.; et al. OBOP: Ontology-Based Ontology Population. http://purl.org/net/obop. Accessed: July 2025
And
Rutesic, P.; et al. OntoSoft Application Models. https://github.com/ontosoft/onto-app-models, 2025. Accessed: July 2025.)
Please let us know if you require that these links be moved into footnotes. Thanks in advance.
Best regards,
authors
Reviewer 4 Report
Comments and Suggestions for AuthorsThis article focuses on the design of multifaceted web applications encoded in knowledge graphs, where all user interaction is treated as a business process. The topic is current and relevant, with clear utility for scientific communication.
Structurally, the manuscript is well organized; however, I believe the abstract should be revised to present the main conclusions more explicitly.
The Results section is mostly descriptive and conceptual, without metrics, comparisons with existing tools, user studies, or data on performance, productivity, or the quality of the information produced. It also does not include an analysis of threats to validity. This section can therefore be substantially improved.
It would be important to add:
A validation benchmark (e.g., with pySHACL) applied to the output graphs, including processing times per form and server throughput.
An objective comparison with at least one SHACL-form tool (e.g., ULB/CSIRO) and with Schímatos on equivalent tasks.
A clearer explanation of how the data entered in the web pages are stored.
Regarding the figures:
In Figure 2, the step “This subprocess is repeated for every existing menue. The list of menues is used to configure the loop” should display a loop symbol, in accordance with BPMN notation.
Figure 6 is difficult to read, and the graphical representation of the flows should therefore be revised and improved.
In addition, the manuscript contains orthographic and typographical errors that must be corrected:
“knowlege”
“menue/menues”
“Shaape”;
In conclusion, the article should be published, but only after these revisions have been implemented.
Author Response
Dear reviewer,
Thank you for your positive feedback and valuable comments. We have carefully addressed those points in the revised manuscript as follows:
Comments1:
Structurally, the manuscript is well organized; however, I believe the abstract should be revised to present the main conclusions more explicitly.
Response1:
We changed the abstract as you suggested to provide a concise explanation of the main conclusions and novelty of our approach.
Comments2:
The Results section is mostly descriptive and conceptual, without metrics, comparisons with existing tools, user studies, or data on performance, productivity, or the quality of the information produced. It also does not include an analysis of threats to validity. This section can therefore be substantially improved.
Response2:
The evaluation and automatic generation of models is planned to be part of a subsequent paper. To make it clear, we changed the discussion of the paper to clarify that the empirical evaluation with numerical results will be carried out in several categories and is planned for a subsequent paper. We are still in the process of generating test data for the evaluation part.
Comments3:
It would be important to add:
A validation benchmark (e.g., with pySHACL) applied to the output graphs, including processing times per form and server throughput.
Response3:
A validation part with pySHACL is included in the model. However, the definition of the validation benchmark has not yet been provided because the processing time consists of multiple times for individual forms in the multi-form application. As can be seen in the user documentation of the project in the git repository, the application can contain one more screen at the end of the application to start the validation and to get the result in the last screen of the application.
https://github.com/ontosoft/onto-app-models/blob/main/documentation/OntoUI-documentation.pdf
Comments4:
An objective comparison with at least one SHACL-form tool (e.g., ULB/CSIRO) and with Schímatos on equivalent tasks.
Response4:
As mentioned the comparison to other tools and validation benchmark has not yet been provided because our processing time consists of multiple times for individual forms in the multi-form application, and we now clarified additionally in the "Discussion" section that the evaluation will be part of a subsequent paper. We are currently working on this.
Comments5:
A clearer explanation of how the data entered in the web pages are stored.
Response5:
Now we clarified in the section that describes the architecture that we use an in-memory triple store to save output knowledge graphs. Additionally, it is explained that the Python RDFLib library is currently used in the prototype for this purpose.
Comments6:
Regarding the figures:
In Figure 2, the step “This subprocess is repeated for every existing menue. The list of menues is used to configure the loop” should display a loop symbol, in accordance with BPMN notation.
Response6:
We added the multi-instance marker with three short vertical lines as a loop symbol in accordance with BPMN notation.
Comments7:
Figure 6 is difficult to read, and the graphical representation of the flows should therefore be revised and improved.
Response7:
We have changed Figure 6 by correcting typos in prefixes and changing the positions of text labels to make it more readable. Additionally, we added necessary prefixes in listings and in Figure 6.
Comments8:
In addition, the manuscript contains orthographic and typographical errors that must be corrected:
“knowlege”
“menue/menues”
“Shaape”;
Response8:
The typos you found and some other typos in images have been corrected.
Best regards,
authors
Round 2
Reviewer 1 Report
Comments and Suggestions for AuthorsThe authors have satisfactorily addressed the provided comments. Below are some points to address before the publication.
- Typographical Error: E.g, "Therefore, our system can be represented in Algorith 1 as two procedures...". Proofread the manuscript.
- End the "While" function/declaration in Algorithm 1.
- Introduce relevant and recent studies to back up most of the mentions/claims in the Discussion (Section 6) and Results (Sections 5.1-5.3)
Author Response
Dear reviewer,
Thank you for your additional remarks. We make corrections in the following way:
Comment1:
- Typographical Error: E.g, "Therefore, our system can be represented in Algorith 1 as two procedures...". Proofread the manuscript.
Response 1:
These types are corrected together with other typos which were found.
Comment2:
End the "While" function/declaration in Algorithm 1.
Response2:
The algorithm is corrected accordingly.
Comment3:
Introduce relevant and recent studies to back up most of the mentions/claims in the Discussion (Section 6) and Results (Sections 5.1-5.3)
Response3:
We introduced additional relevant studies regarding ontology based form generation in the Discussion Section (and Recent work section). Those tools are namely SHAPEness and ActiveRaUL
Additionaly, we added reference to a work with the use of reasoning with SHACL validation (Results section).
At the suggestion of another reviewer, the table with the comparison analysis is once again added with our prototype within. Discussion is accordingly extended with this analysis.
Best regards,
authors
Author Response File: 
Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for AuthorsMost of my comments have been answered.
I recommend the authors add a new table (Table 2) in the result section, like Table 1. The compliance of software tools for HTML form generation with the specified requirements
In Table 2, authors need to state which RQs have been answered by their approach. And, explain in 2-3 paragraphs how each RQ has been answered?
Author Response
Dear reviewer,
Thank you for your additional remarks. We make corrections in the following way:
Comment1:
I recommend the authors add a new table (Table 2) in the result section, like Table 1. The compliance of software tools for HTML form generation with the specified requirements
In Table 2, authors need to state which RQs have been answered by their approach. And, explain in 2-3 paragraphs how each RQ has been answered?
Response1:
We included a new table (Table 2) in the Discussion section with an additional row for our “onto-app-model” solution. In the subsequent paragraphs, we explained how the RQs have been answered. To make it more complete, we introduced references to two additional software solutions that have similar functionalities. Those are SHAPEness and ActiveRaUL. Additionally, we divided the research questions into six categories, as SHAPEness included form applications implemented in the Eclipse Rich Client Platform, which are not web-based forms. Nevertheless, those tools implement SHACL shapes.
Best regards,
authors
Author Response File: Author Response.pdf
