An Automated Framework for Prioritizing Software Requirements
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsManuscript Strengths:
The manuscript addresses a relevant problem in software requirements engineering by proposing an automated framework for prioritization.The use of BERT embeddings for natural language processing and K-Means clustering for grouping requirements is a strong technical choice. The study is well-structured, with a logical flow from problem identification to methodology and results. The RALIC dataset provides a recognized benchmark for evaluating the proposed approach. The Mean Absolute Error (MAE) metric is used to measure prioritization accuracy quantitatively.
Limitations & Areas for Improvement in the manuscript:
The literature review does not sufficiently differentiate the proposed method from existing approaches. It needs more explicit comparison with alternative RP techniques, such as StakeRare, SRPTackle, or Genetic K-Means, is needed.
The justification for selecting K-Means over alternative clustering techniques (e.g., DBSCAN, hierarchical clustering) is not well-explained. A comparative analysis is much needed.
The methodology presented lacks technical depth in explaining the ranking mechanism within clusters and how requirement importance levels are weighted.
The evaluation relies solely on MAE, whereas additional metrics such as Precision-Recall, F1-score, or NDCG could provide a more comprehensive assessment.
There is no comparison with traditional or hybrid prioritization methods (e.g., AHP, MoSCoW, or multi-criteria decision-making techniques).
The generalizability of the framework beyond the RALIC dataset is unclear, and no real-world case study is provided.
While reducing stakeholder involvement improves objectivity, it also removes domain expertise and contextual insights that might be crucial in complex projects.
The manuscript does not address how the framework would handle evolving requirements, conflicting priorities, or dynamic project constraints. A discussion is needed on this aspect too.
A hybrid approach, integrating limited stakeholder feedback at critical stages, could improve practical applicability.
Include a real-world case study to demonstrate the framework’s practical applicability.
Address ethical and practical challenges in requirement prioritization automation and explore hybrid solutions.
Comments on the Quality of English LanguageThe manuscript needs to be proofread by native speakers.
Author Response
Our response letter is attached.
Author Response File: Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsThis paper proposes an automated framework for prioritizing software requirements but is judged to fall short of meeting the fundamental standards of an academic paper. The primary reasons for this assessment are as follows:
First, the paper fails to rigorously explain the operational principles of BERT and K-Means algorithms with mathematical formulations, leading to insufficient theoretical validity. Additionally, the lack of detailed explanations regarding the model's implementation and parameter settings raises concerns about reproducibility.
Second, the interpretation of experimental results and the visualization materials (e.g., graphs and tables) are overly simplistic, lacking sufficient data analysis and statistical evidence to support the claims, which significantly undermines the paper's persuasiveness.
Third, the comparison with existing studies is inadequate, and the proposed methodology's originality and academic contribution are not sufficiently substantiated.
For these reasons, the paper lacks the necessary academic depth and reliability, and a decision to reject is considered appropriate.
Author Response
Please check the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for AuthorsI like the work very much. What should be improved (expanded) is adding an example with a few dozen (max 20-30) requirements illustrating the operation of the classical and automatic method. In particular, it would be interesting to show the assignment of requirements to clusters and assess their similarity - which is crucial in the method under discussion.
The example will illustrate the operation of the method and significantly enrich the article.
Author Response
Please check the attachment.
Author Response File: Author Response.pdf
Reviewer 4 Report
Comments and Suggestions for AuthorsSpecific Comments
-Page 1: There is a sudden shift in from RE to RP? What exactly are the authors trying to address, RE or RP?
-Page 2: What are the RP techniques? What is the meaning of RP? Define acronyms before using them in the manuscript.
-Page 3, line 103: Delete the extra “]” in the statement, “Reyad, Omar, et al. [8] ]”
-Page 3, Line 114: The statement, “However, its focus on a single…” is ambiguous. Kindly recast.
-Page 3: What is RALIC? Define acronym before using it.
-Page 4; Line 156: Provide the reference for the RALIC Dataset as the current reference does not refer to the dataset. Provide a description of the RALIC Dataset and provide the justification for its choice.
-Page 4-5; The information provided in section 3.3 to 3.8 is more generic and redundant information without providing how they are applicable to the manuscript. Authors should expunge redundant information and align their explanation in the context of the subject matter.
-Page 6: The proposed method lacked depth in its description. Authors should provide a detailed description of the proposed method showing how the stated problem is addressed by the proposed methodology and system setup. Most of the discussions in section 4 are repetitions of section 3.4 to 3.8. Provide a detailed description of the component/structure of adopted machine learning model.
-On Page 2, the authors stated that, “This paper introduces a machine learning framework designed to alleviate the workload on stakeholders and enhance the efficiency of prioritizing large and complex sets of requirements. The framework aims to streamline the prioritization process by automating the comparison tasks, thus addressing the critical scalability issues identified in existing RP methods.” On the contrary, the result analysis and discussion did not provide convincing evidence of how this stated objective was achieved. How were the critical scalability issues identified in existing RP methods addressed by your result? Compare your result with existing RP methods highlighting in specific terms based on the result of how your proposed method faired with exiting methods.
Comments for author File: Comments.pdf
The entire manuscript needs to be re-written
Author Response
Please check the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
Comments and Suggestions for AuthorsProof reading is necessary for improving overall the language
Comments on the Quality of English LanguageA proof read is necessary for removing the grammatical mistakes
Author Response
Q: A proof read is necessary for removing the grammatical mistakes
We are grateful for the positive comments from the reviewer. We have proofread the paper. Corrections made to the paper are shown in blue.
Reviewer 3 Report
Comments and Suggestions for AuthorsNow it looks good!
Author Response
Q: Now it looks good!
Thank you for this positive comment.
Reviewer 4 Report
Comments and Suggestions for AuthorsSpecific Comments
- Page 17 of 22. Table 7 write-ups are blurry and not readable. Authors should improve the readability of Table 7.
- Similarly, Page 18 of 22, The write-ups in Table 8 are blurry and appear to be censored. Authors should improve Table 8 and make the write-ups readable.
Overall, the manuscript content has improved.
Comments for author File: Comments.pdf
Author Response
Q:
- Page 17 of 22. Table 7 write-ups are blurry and not readable. Authors should improve the readability of Table 7.
- Similarly, Page 18 of 22, The write-ups in Table 8 are blurry and appear to be censored. Authors should improve Table 8 and make the write-ups readable.
Overall, the manuscript content has improved.
A: We appreciate the reviewer's positive comment. We agree with the reviewer on the suggested corrections and have added both tables of appropriate quality.
Author Response File: Author Response.pdf