The expansion of image information using ontologies to enhancing the efficiency of data retrieval tasks

Dear Editor,

After reviewing the manuscript titled "The expansion of image information using ontologies to enhance the efficiency of data retrieval tasks," I would like to share my perspective on its suitability for publication. In my opinion, the paper does not meet the necessary standards in its current form.

The contribution of the work is insufficiently contextualized to emphasize its relevance and novelty within the field. While the authors state that the primary objective is to enhance data retrieval efficiency through the use of ontologies, the content largely focuses on basic aspects of ontology construction. This emphasis appears misaligned with the stated goals of the paper.

Furthermore, the manuscript devotes substantial attention to the technical details of the library created, including the specific libraries and scripts utilized. This approach resembles a user manual for the code rather than a well-rounded scientific study. While some methodological explanation is expected, the disproportionate focus on implementation details detracts from the conceptual contributions and theoretical insights that are critical for impactful work in this area.

Although the core idea of the paper has potential, the current presentation lacks the depth, clarity, and alignment needed to establish its significance. I believe the work requires substantial rethinking and refinement before it can be considered for publication.

Thank you for the opportunity to review this manuscript.

Sincerely

The paper presented an improved approach that uses a custom-designed ontology specifically created to semantically describe image information in web documents. This ontology enables a richer and more nuanced representation of image data, facilitating better understanding and retrieval. In the first phase of research they focused on animals, based on hierarchical division of animals according to the encyclopedia BIOpedia. Additionally, they envision integrating these semantic descriptions with image information derived from deep learning techniques, creating a comprehensive and unified description. The proposed approach aims to significantly enhance the accuracy and efficiency of data retrieval in both databases and web search engines.

Regarding the paper:
(1) originality: the topic of enriching the image information using ontologies is already presented previously, and the authors mentioned that in the related work section.
But for example, they didn’t mention that there is a previous paper also constructed animal ontology for image retrieval:
https://link.springer.com/article/10.1007/s00530-007-0099-4
So, authors should mention clearly what are the differences between their work, and previously presented works.
(2) significance of the results: good
(3) quality of writing: good

Regarding the “Long-term stable URL for resources”:
(A) and (B) The data files are well organized but don’t contain a README file, which makes it difficult to assess the data, and replicate the experiment. Authors should add README file that includes detailed steps to how to replicate the experiment.
(C) and (D) the authors added an offline copy of all files.

Writing comments for authors:
(1) The authors mentioned “deep learning techniques” in the abstract and introduction and didn’t use any of them through the implementation!!! Where is the section of using these techniques? If not used, authors should remove this word from abstract and introduction.
(2) Refer to your achieved results in the abstract.
(3) Add the reference the first time you mention it. For example protégé (page 2 – line 14), model (page 2 - line 50), and ontology definition (page 3 – line 35) ……..
(4) Explain what the semantic gab is (page 2 - line 39)
(5) meanings and meanings (page 3 - line 40)
(6) Better to rename section 4 to “design of proposed ontology”
(7) Add summary paragraph at the end of related work section to differentiate between proposed method and previous works.
(8) In section 4.3 mention the total number of individuals.

The paper's proposed method involves the use of a custom-designed ontology specifically created to semantically describe image information in web documents. This ontology enables a richer and more nuanced representation of image data, facilitating better understanding and information retrieval. This proposed approach aims to significantly enhance the accuracy and efficiency of data retrieval in both databases and web search engines. In addition, the author proposes a method for building an ontology and experiments with animal data enriched for the ontology with images. The paper contributes to the construction and testing of an ontology of the animal data domain. The ontology design process is illustrated by Protege, the application is tested by combining machine learning techniques with Python libraries.
(1) Originality: The article is guaranteed to be original
(2) Significance of the results: Effective and reliable experimental results
(3) The resources provided seem to be sufficient to replicate the experiments. The data file is well-organized.
(4) The data provided resources complete and the data is provided in the paper complete.
(5) Quality of writing: The quality of the article as well as the language of the plow should be edited according to the content of the comment below.

The author needs to review and make major revisions with the following comments:
1. The Related Work section mentions TBIR, CBIR, and SBIR but does not focus on the main content of the paper. In this section, it is necessary to analyze related works that need to be referenced to develop or improve to expand the ontology for experimentation.
2. The ontology design process with image attribute extension for the animal data domain should be presented in more detail from the ontology framework design for the animal domain to clarify the contribution of the paper.
3. The author should revise the explanation of algorithms. In the algorithms, the author should give the algorithm, present the algorithm idea, and the input and output data. Then evaluate the complexity (if any) to avoid long explanations, for example in section 5.6:
- The first extraction example, found on lines 5-7, uses the "soup" attribute of the object (a reference to an object already parsed by BS4) to retrieve the title of the web document. If the "title" tag is not present in the document, the value "None" will be assigned. In the second example on line 9-14, we use the "find_all" method, which can retrieve an array of all the tags found in the document, in our case we have determined that we are searching for "p" tags, which usually contain text in web documents. The result of this method is returned in the "paragraphs" variable, which is then used in the "for" loop to display the text of each element of our paragraphs array, using the "text" attribute,
4. At the end of section 5.6, the author's explanation is long but does not focus on the main content, making it difficult for readers to understand, for example:
- “Another cause of errors, as already mentioned in the introduction of this subchapter, was the fact that the animal was identified in the image information where it was not present, and these were mostly cases of images of company logos and promotional graphics of the websites that published the image information”
5. The part of the Ontology design is done visually using Protégé, the application construction experiment uses Python libraries, so the author needs to present the experimental content more clearly.
6. The algorithms should present input and output data.
7. Figures 12 and 13 have the same name, need to check again.
8. figures 4, 5, 6, 9, 10, 11, 12, 13, and 17 should use the English language to design Ontology for the animal domain.
9. The text in images 4, 17, and 21 are quite small and can be adjusted to a larger size.
10. Figure 18 will show a sample image tested from the video which will be more convincing and clear.
11. The English and Slovak languages are mixed in the paper, so the authors should use English for the full text.
12. The paper is presented in an active voice quite a lot, such as:
- We created a class called "properties".
- We chose the "Individuals" tab.
- By pressing the "Create Instance" button, we have created an individual that is uniquely defined as "Asserted".
- We called the new individual.
- After creating the individual instances, we can now describe the different classes of species.
- we chose to use a graphical element available in the CustomTkinter library, named "CTk.Tabview", which is convenient precisely because it allows us to insert multiple tabs that are created when displaying multiple results from the ontology file that contain the refined descriptions of the animals
13. It is necessary to clearly present the experimental data set, in table 2 the column “Number of data on the web” is presented, the data in this column is quite modest.
14. The author should add more references to compare, analyze, and evaluate the results, thereby finding solutions to design ontology for animal domain and limitations.
15. The contributions are still general, the author should clearly present the main contributions of the paper.
16. Experimental section 5 is too long, and explaining the algorithm's lines is unnecessary. The experimental results need to analyze the values obtained.
17. The conclusion is too long, so it is necessary to present the outstanding results that the article has contributed and the remaining limitations to have further direction.