Semantic Augmentation in Chinese Adversarial Corpus for Discourse Relation Recognition Based on Internal Semantic Elements

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The authors' interest in their work is in the domain of natural language processing, explicitly dealing with adversative complex sentences. To address the problem, the authors suggested incorporating linguistic semantic information into discourse relation recognition and constructing a Semantic Augmented Chinese Discourse Corpus. The authors have used classification models with deep learning methods to prove the effectiveness of internal semantic element features and the applicability of the SACA corpus.

• The paper is well constructed.
• The contributions are presented in the paper and well described.
• The authors presented a good overview of the state of the art.

The paper can be considered for publication in its current state, but changes must be made to the final version.

 

• The outline section needs to be added to the introduction section.
• The author should start with a small introduction in sections two and three.
• In Equation 1, there is no description of the variables used or what each refers to.
• The author should add more details when discussing the results since they have used different methods. It would be better if they spoke about each one.
• Are there any AI techniques that detect user behaviour instead of statistical techniques?
• Natural language processing with Deep learning could detect user favourability; why did the authors not make any statements about this matter?

Author Response

Dear Reviewer,

 

Thank you for your thorough review and constructive comments on our manuscript Semantic Augmentation in Chinese Adversarial Corpus for Discourse Relation Recognition Based on Internal Semantic Elements. We appreciate the opportunity to enhance our work and clarify the points raised. Below are our responses to the comments and the corresponding changes we have made to the manuscript.

 

• Comment 1: The outline section needs to be added to the introduction section.

 

As per your suggestion, we have added an outline section at the end of the introduction. We appreciate this recommendation as it has helped provide a clearer roadmap of the paper's structure, thus aiding readers in navigating our study more effectively.

 

• Comment 2: The author should start with a small introduction in sections two and three.

 

We agree with the feedback that sections two and three lacked a brief introductory part. We have now included a small introduction at the beginning of each section to set the context and link the sections coherently with the rest of the paper.

 

• Comment 3: In Equation 1, there is no description of the variables used or what each refers to.

 

We apologize for the oversight in not describing the variables in Equation 1. We have now amended this by adding a detailed description of each variable immediately following the equation. This clarification should aid in understanding the mathematical model used in our research.

 

• Comment 4: The author should add more details when discussing the results since they have used different methods. It would be better if they spoke about each one.

 

Thank you very much for suggesting that more details should be added when discussing the results. Following your advice to discuss each method in detail, we have added Section 5.3, which elaborates on the different methods used. Additionally, in Section 5.4 where the results are presented, we have provided more detailed explanations for Table 5 and Table 6. This clarification should help readers understand the models used in our research and enhance the overall clarity of the manuscript. Thank you for highlighting this issue.

 

• Comment 5: Are there any AI techniques that detect user behaviour instead of statistical techniques?

 

Thank you very much for your comment. You mentioned that some AI techniques can be used to detect user behavior instead of statistical techniques. Indeed, we have utilized such techniques but failed to mention them in the paper, for which we apologize. We have used these techniques to identify the annotators' gender, grade, major, the time period of annotation, and the duration of engagement to infer their annotation preferences. We have added this discussion at the beginning of Section 3.4.2.

 

• Comment 6: Natural language processing with Deep learning could detect user favourability; why did the authors not make any statements about this matter?

 

Thank you for your insightful question regarding the absence of a discussion on using deep learning to detect user favorability in our paper. We apologize for not addressing this in our manuscript. As you rightly pointed out, deep learning can indeed be instrumental in discerning user preferences, a method we have incorporated to some extent in user selection processes, as described in response to Comment 5.

 

While our study utilized AI technologies for certain tasks, we recognize that natural language processing (NLP) coupled with deep learning could enhance our ability to annotate or interact more effectively within AI interfaces. Regrettably, our annotation processes did not employ these specific techniques, which explains their omission from our discussion.

 

We are truly grateful for your suggestion and will certainly consider integrating NLP with deep learning techniques in future work to detect user favorability better. Your feedback is invaluable to us, and we thank you once again for bringing this to our attention.

 

We hope these revisions satisfactorily address your concerns and significantly improve the manuscript. We are grateful for your insights, which have undoubtedly strengthened the paper.

 

Thank you once again for your constructive critique and valuable time spent on our manuscript.

 

Best regards,

The Corresponding Author

May 2, 2024

 

Reviewer 2 Report

Comments and Suggestions for Authors

The manuscript presents interesting linguistic concepts, detailed annotation and introduction for new Chinese dataset and an interesting model that beats multiple other models for CADRR task. 

I would suggest that the authors to revise the presentation of some parts of the paper. For example, 

1. Rewrite the abstract as the current one is very confusing to the readers.

a. A lot of abbreviations appear in the abstract before its first time appearance of the full name and without any citations, such as PDTB, HIT-CDTB. I understand that it's hard to explain all of them in a short abstract. I would suggest describing them using more plain languages in the abstract to make everything more clear.

b. rephrase the abstract to highlight the paper's novel contribution more.  The current abstract is not clear enough for understanding what the paper's highlight is.

2. Add more descriptions when you explain the concepts/tables. 

a. For example, in equation (1), what do the variables stand for?

b. In Table (2), what are the definitions of Overall Sense and internal semantic elements.

It seems captions and/or more descriptions in the texts are needed for the tables.

3. The author should review everywhere else to see if there is any concept that hasn't been explained briefly and clearly before it appears, which are fundamental barriers for the readers to understand the paper. 

Overall, it's an interesting paper with sufficient work to be published. 

Author Response

Dear Reviewer,

 

Thank you for your constructive feedback on our manuscript, Semantic Augmentation in Chinese Adversarial Corpus for Discourse Relation Recognition Based on Internal Semantic Elements. Based on your valuable comments, we appreciate the opportunity to improve our work. Below, we address each point raised and outline the revisions we have made to the manuscript accordingly.

 

• Comment 1: Rewrite the abstract as the current one is very confusing to the readers.

 

1. Abbreviations and Citations: You rightly pointed out that the abstract contained several abbreviations without their full forms or proper explanations, making it difficult for readers unfamiliar with the topic. We have revised the abstract and either defined all abbreviations at their first mention or replaced them with plain language descriptions. For instance, "PDTB" and "HIT-CDTB" have been replaced with their full names and a brief context to ensure clarity.

 

1. Highlighting Novel Contributions: We have also rephrased the abstract to highlight our paper's novel contributions better. Now, the abstract explicitly states the main advancements our work offers over existing methods in the field of DRR, emphasizing the practical implications and improvement our dataset presents and the CADRR task defined in the paper.

 

• Clarification of Concepts and Tables

 

1. Equation (1) Clarifications: In response to your suggestion, we have added a detailed description of each variable in Equation (1) directly below the equation itself. This includes a brief yet comprehensive explanation of the variables and their significance in the context of our proposed model.

 

1. Table (2) Definitions: We have updated the caption of Table (2) to include definitions of "Overall Sense" and "Internal Semantic Elements."

 

For the other tables in the paper, we have added necessary explanations to the captions of some tables, and for others, we have included additional descriptions in the main text. Thank you very much for your suggestions.

 

• Review and Explanation of Fundamental Concepts

 

Following your advice, we have thoroughly reviewed the manuscript to ensure that all fundamental concepts and terms are introduced and explained clearly before they are used extensively. This includes revising sections that previously assumed reader familiarity with certain terms or methodologies, ensuring the manuscript is accessible to a broader audience.

 

In addition to these specific changes, we have made several stylistic and grammatical improvements throughout the manuscript to enhance its readability and flow. These adjustments contribute to a clearer and more compelling presentation of our research.

 

We believe that these revisions have significantly improved the manuscript, addressing the concerns raised in your review. We are grateful for the insights provided, which have undoubtedly strengthened the paper.

 

Thank you for your attention and support throughout this process.

 

Best regards,

The Corresponding Author

May 4, 2024

 

Reviewer 3 Report

Comments and Suggestions for Authors

Very interesting article. Original subject matter. In developing international (social and economic) relations, the article should be considered very important.

The introduction should indicate the research problem (it is included in the article).

This is not mandatory, but it is worth writing down research hypotheses. This will increase the clarity of the article.

The data source used requires a broader description. It's about the international reader.

In table 3, column 3 in the heading reads: 'Percentage'. Therefore, the '%' symbol should not be used for quantitative data written below.

Figure 3 is a table. The description should be changed. The interpretation of the data contained in this table should be expanded. The note also applies to the data contained in the remaining tables.

There is no discussion of the results - no reference to other research/scientific discoveries.

In the conclusion, you should indicate what the article's contribution to science is. On a practical level, it is worth pointing out a context beyond China.

It is worth writing about research limitations, if any.

Author Response

Dear Reviewer,

 

Thank you for the comments regarding our manuscript, Semantic Augmentation in Chinese Adversarial Corpus for Discourse Relation Recognition Based on Internal Semantic Elements. We greatly appreciate the insights and constructive criticism, which have helped us improve the quality and clarity of our paper.

 

Below, we address each comment point-by-point and describe the changes we have made to the manuscript accordingly.

 

• Comment 1: The introduction should indicate the research problem (it is included in the article).

 

As suggested, we have revised the introduction to indicate the research problem more clearly. This amendment helps to better set the stage for our study’s objectives, which we have now explicitly linked to the broader field of international relations.

 

• Comment 2: This is not mandatory, but it is worth writing down research hypotheses. This will increase the clarity of the article.

 

We agree with the reviewer’s suggestion that research hypotheses enhance the clarity of the article. Based on our initial research questions, which are outlined at the end of the revised Section 2.2, we have now formulated and included specific hypotheses. We hypothesize that internal semantic elements are very important for the recognition of adversative sentences.

 

• Comment 3: The data source used requires a broader description. It's about the international reader.

 

Thank you for your valuable comment regarding the description of our data sources. To better accommodate international readers and provide a more comprehensive context, we have expanded the description of our data sources in Section 3.3.1 of the revised manuscript. We believe this enhancement will make our study more accessible and understandable to readers from diverse backgrounds. We appreciate your feedback as it has significantly contributed to improving the quality of our paper.

 

• Comment 4: In table 3, column 3 in the heading reads: 'Percentage'. Therefore, the '%' symbol should not be used for quantitative data written below.

 

Thank you for pointing out the redundancy in the heading of column 3 in Table 3. We have addressed this issue by changing the heading to "Contribution," which not only avoids redundancy but also ensures that the representation of quantitative data is more accurate and aligns with standard data presentation practices. We appreciate your attention to detail, which has helped enhance the clarity and quality of our presentation.

 

• Comment 5: Figure 3 is a table. The description should be changed. The interpretation of the data contained in this table should be expanded. The note also applies to the data contained in the remaining tables.

 

Thank you very much for your attention to Figure 3. Indeed, Figure 3 resembles a table, which is due to our omission of color information in the annotation, a point you also highlighted. We have added a note about color information in the annotation.

 

• Comment 6: There is no discussion of the results - no reference to other research/scientific discoveries.

 

We have added a new paragraph at the end of Section 5.4 that discusses some references in other research/scientific discoveries. As this corpus is a new dataset, no previous research has been conducted on it. Our task, known as CADRR, is an initiative we have proposed based on the DRR task. To validate the reasonableness of our task and the applicability of our corpus, we compared it with the currently popular IDRR task, which is also based on DRR.

 

• Comment 7: In the conclusion, you should indicate what the article's contribution to science is. On a practical level, it is worth pointing out a context beyond China.

 

In the revised conclusion, we clearly state our article's contribution to the scientific community and elaborate on its practical implications, particularly outside China. The methods and theoretical frameworks are broadly applicable and can be extended to other languages and cultural contexts, thereby supporting the recognition and analysis of complex discourse relationships worldwide.

 

• Comment 8: It is worth writing about research limitations, if any.

 

We greatly appreciate your feedback. In response, we have included a paragraph addressing research limitations at the end of Section 6. Our study's limitations encompass the scope and diversity of the corpus, which presently centers primarily on specific types of adversative relations. Moreover, while our method demonstrates strong performance on the current dataset, further validation is necessary to assess its generalizability and effectiveness across diverse linguistic contexts. Thank you for bringing these important considerations to our attention.

 

We trust that these revisions adequately address the comments and significantly enhance the manuscript. We are grateful for the opportunity to improve our work and thank you for your continued consideration.

 

Thank you once again for your attention and guidance.

 

Best regards,

The Corresponding Author

May 5, 2024

Reviewer 4 Report

Comments and Suggestions for Authors

The authors of this manuscript proposed utilizing linguistic semantic information into the recognition of discourse relations and the creation of a Semantic Augmented Chinese Discourse Corpus (SACA) filled with adversative complex sentences. They introduced a quadruple structure (P, Q, R, Q_β) to represent key semantic elements within these sentences, highlighting the adversative relationship primarily formed by the semantic opposition between Q and Q_β. P is identified as the premise, while R is the adversarial reason. The study attempts to link sense classification and internal semantic elements. Also, it introduces a new task, the Chinese Adversative Discourse Relation Recognition (CADRR). Finally, the manuscript details the development of classification models using deep learning techniques, demonstrating the efficacy of internal semantic elements and the practicality of the SACA corpus. The proposed work outperforms pre-trained models by integrating internal semantic elements.

 

This is an interesting manuscript of an interdisciplinary nature, that lies in the intersection of linguistics with deep learning.

 

While there are some stylistic language mistakes, such as joining a references to words (e.g. analysis[1], reasoning[2], etc…), nevertheless, the manuscript is well-written and comprehensible.

 

The main strengths of the manuscript relate to its orderly structure, comprehensibility (even though it tackles sentence structures in Chinese), and organization. Apart from the usual sections found in manuscripts (Introduction, Methodology, Results, and Conclusions), Section 3, “The Corpus Construction” is interesting to read and grasp. The various tables, figures and mathematical formulas also aid in comprehension. The results subsection shows excellent performance, based on calculated values of various metrics. The references are rather recent and adequate.

 

With regards to weaknesses, I cannot say that I see a weakness in this work. My only comment would relate to some acronyms which are given without an explanation (e.g. PDTB?!, RST?!, etc…).

 

Also, while I do like this manuscript, I am not entirely sure of its suitability to MDPI’s Electronics. I believe that other MDPI journals could be more suited for such a manuscript (e.g. Computation). But of course, such a decision remains at the discretion of the Editor.

 

Ultimately, I would like to congratulate the authors of this manuscript on carrying out such research work of an interdisciplinary nature!

Comments on the Quality of English Language

Given above.

Author Response

Dear Reviewer,

 

Thank you for the detailed and encouraging feedback on our manuscript entitled Semantic Augmentation in Chinese Adversarial Corpus for Discourse Relation Recognition Based on Internal Semantic Elements submitted to MDPI's Electronics. We appreciate your insights and suggestions which have helped us improve the quality and clarity of our work. Below, we respond to the comments point-by-point and describe the revisions we have made to the manuscript.

 

• Comment 1: There are some stylistic language mistakes, such as joining references to words (e.g., analysis[1], reasoning[2], etc…).

 

We acknowledge the stylistic errors noted, particularly the incorrect joining of references to words (e.g., analysis[1], reasoning[2]). We have carefully revised the manuscript to correct these mistakes and ensure that all references are properly formatted according to the journal’s guidelines.

 

• Comment 2: Some acronyms are given without an explanation (e.g., PDTB?!, RST?!, etc.…).

 

We appreciate your observation regarding the unexplained acronyms (e.g., PDTB, RST). Thank you for bringing the issue of unexplained acronyms to our attention. We apologize for any confusion caused by this oversight. In the revised manuscript, we have now ensured that each acronym is fully spelled out at its first occurrence, and we have added necessary annotations throughout the document to aid reader comprehension. This should make the text more accessible and enhance understanding of the discussed concepts.

 

We believe that the revised manuscript addresses the concerns raised during the review. We hope that the revisions meet your expectations.

 

Thank you once again for your detailed comments and for considering our work.

 

 

Best regards,

The Corresponding Author

May 4, 2024