Research on a Web System Data-Filling Method Based on Optical Character Recognition and Multi-Text Similarity

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This paper presents a web system data-filling method based on OCR technology. Overall, the paper lacks a lot of key elements that should be included in a technical article like this. The reviewer has the following questions and comments, which need to be addressed in the revised version of the manuscript. The current version is not suitable for publication.

 

1)    In Section 1 (Introduction), the literature review is not enough. Please perform an exhaustive literature review on this domain and find the gaps. Then, write your unique contribution based on the gaps. It is extremely hard to believe that nobody has done any kind of data-filling method from handwriting, as handwriting recognition was one of the first stages in machine learning research.

 

2)    In Figure 2, please present the architecture in detail, such as the dimensions of each layer and dropouts. Please also perform a proper hyperparameter tuning and include it in the paper as well.

 

3)    In section 4 (Experimental result and analysis), the experiment has been conducted on your dataset. Is there no available dataset? Please search carefully. If it is really not available, then publish or give a link to your own dataset for verification along with this paper.

 

4)    In section 4, what is the reason for choosing only 80 images in the dataset?

 

5)    How do the authors ensure that 40 images are enough to train the model? Properly show the loss function in detail to convince the readers that the model is trained.

 

6)    Figures 11 and 12 are not explained properly.

 

7)    Please show an exhaustive comparison with current state-of-the-art methods.

Author Response

Dear Reviewer,

Please see the attachment.

Best regards,

Hailu Su

Author Response File: Author Response.docx

Reviewer 2 Report

Comments and Suggestions for Authors

The paper proposes a method to automatically fill in complex form images in web systems by combining OCR technology and Levenshtein multi-text similarity. The authors conducted experiments and found that the proposed method achieved a filling accuracy of over 90% for complex form images. However, the study did not report any data, and the authors declared no conflicts of interest. The paper also discusses related technologies and research, such as OCR recognition technology and text similarity algorithms. The authors concluded by emphasizing the potential for further optimization of the method to adapt to different image forms and real-time updates of database tables corresponding to fields.

There some comments:

(1)Can you give the traditional methods of data uploading in web systems?

(2)Can you explain how technology contributes to the data filling method proposed in the article?

(3)What are the evaluation criteria used to measure the accuracy of the system's filling method?

(4)How were the experimental results used to verify the effectiveness of the proposed method?

Author Response

Dear Reviewer,

Please see the attachment.

Best regards,

Hailu Su

Author Response File: Author Response.docx

Reviewer 3 Report

Comments and Suggestions for Authors

Paper deals with important task. It has a great practical value. The scientific novelty must be clarified. It has a logical structure. The experimental section must be improved. The proposed approach is logical. The result section must be improved.

Suggestions:

1.     The introduction section should be expanded with a clearer description of the motivation and scientific novelty of the study.

2.     It would be good to add the remainder of this paper.

3.     Figure 3 shows the Functional Structure Diagram and Figure 4 shows the Identification process diagram. It is necessary to describe in detail each component of these diagrams and the connections that exist between them.

4.     In section 3, it is necessary to highlight the methodology and components proposed by the authors. And emphasize the advantages that they provide.

5.     Authors should provide a link to open access repository with the dataset in the Reference section.

6.     It is also necessary to conduct an analysis of the created dataset, to show its completeness, balance, and relevance to the researched problem.

7.     On the basis of which criteria were determined the proportion of irrelevant information and the proportion of similar fields in the image. How time-consuming will such operations be in case of expansion of the validation dataset?

8.     The results shown in Figures 11 and 12 need to be described in more detail. They are based on ratios (3) and (4) respectively, but it is unclear which strings are being compared? There are many fields on the test form, how is the aggregate value obtained?

9.     I also consider it appropriate to expand the introduction section by analyzing ensemble methods of digital image processing for handwriting recognition. In particular, pay attention to such publications doi: 10.1016/j.compeleceng.2021.107111, doi: 10.1109/ACPR.2017.5 and doi: 10.1007/978-3-031-24475-9_56.

10.                       Conclusion section should be extended using:

·        numerical results obtained in the paper;

 

·        limitations of the proposed approach.

Author Response

Dear Reviewer,

Please see the attachment.

Best regards,

Hailu Su

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

Still, the comment (7) has not been addressed.

7)    Please show an exhaustive comparison with current state-of-the-art methods. 

If there are no state-of-the-art methods to compare with, please justify it in the paper.

 

Author Response

Dear Reviewer,

Thank you very much for your valuable feedback and suggestions on our manuscript. We greatly appreciate your time and expertise.

Regarding the modifications you have suggested, we have carefully considered and made the following revisions:

Comments 7: Please show an exhaustive comparison with current state-of-the-art methods. If there are no state-of-the-art methods to compare with, please justify it in the paper.

Response 7: In response to this issue, we have consulted relevant literature and conducted a detailed content comparison. The specific modifications are marked as 5 in the word.

We believe that these modifications have further improved the quality and accuracy of the paper. Once again, we sincerely thank you for your valuable input and professional assessment, as your suggestions are highly important to us.

Best regards,

Hailu Su

Reviewer 3 Report

Comments and Suggestions for Authors

Most of the recommendations have been taken into account.

Author Response

Dear Reviewer,

Thank you very much for your valuable feedback and suggestions on our manuscript. We greatly appreciate your time and effort in reviewing our work. Your insights have been instrumental in improving the quality and accuracy of our paper.

Best regards,

Hailu Su