All reviews of published articles are made public. This includes manuscript files, peer review comments, author rebuttals and revised materials. Note: This was optional for articles submitted before 13 February 2023.
Peer reviewers are encouraged (but not required) to provide their names to the authors when submitting their peer review. If they agree to provide their name, then their personal profile page will reflect a public acknowledgment that they performed a review (even if the article is rejected). If the article is accepted, then reviewers who provided their name will be associated with the article itself.
Thanks to the authors for your efforts to improve the article. I'm pleased to inform you that this version is ready for acceptance. Congrats!
N/A
N/A
N/A
This revised version has addressed all of my concerns, and I believe this paper can now be accepted.
The authors addressed some issues. However, the experiments are not enough. Moreover, new references should be analyzed.
no comment
no comment
no comment
N/A
N/A
N/A
1. In the revised manuscript, the latest baseline used for comparison is the CKGC algorithm in 2022. Furthermore, even compared with CKGC, the proposed FEGNN method obtains trivial advantage, particularly in CTR Prediction. It is crucial to identify the optimization improvements of FEGNN in terms of time complexity, space complexity, or other relevant factors to substantiate its significant superiority over the baseline.
2. The manuscript exhibits a comparatively low proportion of references from the past three years, and the Introduction section fails to adequately review the developments and achievements in the research field. This indicates a lack of systematic and in-depth exploration of the domain in the manuscript. Further modifications are required to address these concerns.
Please revise the work according to the comments. Then it will be evaluated again.
[# PeerJ Staff Note: The review process has identified that the English language must be improved. PeerJ can provide language editing services if you wish - please contact us at [email protected] for pricing (be sure to provide your manuscript number and title) #]
This paper proposes a recommendation framework that integrates knowledge graph and text-based methods to address the challenges faced when selecting suitable machine learning methods for target datasets. In the proposed framework, to obtain richer latent representations, each node is represented not only by descriptive information in the knowledge graph but also by involving its neighborhood information to enhance the representation, capturing more comprehensive higher-order information during entity propagation. I find the ideas presented in the paper interesting and meaningful, but I still have some concerns as follows:
1. The model framework FEGCN proposed in the paper seems to share many similarities with the state-of-the-art baseline model (DEKR), such as using linking, graph multi-hop propagation, and fusion of two embedding representations. Therefore, I believe this paper should be positioned as a modification of the DEKR model. If the authors have taken this approach, it needs to be clearly stated in the paper how it differs from DEKR.
2. There are still some Chinese characters in Figure 2 of the paper, and some symbols are unclear due to deformation.
1. I believe it would be beneficial to test the proposed model on different datasets because the improvement over the DEKR model is not significant, and testing on a single dataset may introduce randomness.
2. The paper could provide an introduction to the formulas of the evaluation functions.
As it stands, the paper lacks innovation. I hope the authors can clearly explain the differences from DERK in the paper and highlight the specific contributions of this paper compared to DERK.
no comment
1. Enhance English writing skills. For example, the phrase "The first-order neighbours of 'MovieLens-20M' are 'Tasks' such as CTR Prediction, Recommender Systems, Knowledge Base. Besides 'Methods' such as KNI, Deep-FM, and KGCN are the first-order neighbours" is confusing. Additionally, the causal relationship in "However, in most knowledge graph-based recommendation methods, each entity is only represented by its name and informative messages carried by are ignored. Therefore, the paper captures the enhancement features for each node by aggregating their first-order neighbour information" is weak.
2. The Introduction section dedicates a significant amount of space to discussing Figure 1. However, we advise to focus on categorizing or discussing the development process of research papers in the field.
3. The research direction of the paper is knowledge graph + graph networks, but there is a lack of introduction to graph networks in the related work section. It is recommended to include relevant works.
4.The piecewise function in Equation 1 needs to be aligned properly.
5.Normalize the notation of the equations. For example, in Equation 2, it is recommended to use different symbols 'e' for the numerator and denominator, and avoid using the same symbol '\phi' for both the input and output before and after normalization.
6.Symbols are missing on lines 207-208.
This study is relatively novel. However, the latest baseline algorithm used is from 2021. More recent baseline algorithms within the past three years should be included.
no comment
All text and materials provided via this peer-review history page are made available under a Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.