Research on Learning Resource Recommendation Based on Knowledge Graph and Collaborative Filtering

Round 1

Reviewer 1 Report

It is a very interesting, up-to-date investigation with empirical evidence. The results are in line with the interests of the student body. Teachers should consider these findings to improve student learning.

It is suggested to add limitations of the study.

Author Response

Dear Reviewer,
Thank you very much for taking the time to review this manuscript. Thank you very much for your feedback and suggestions. Thank you very much for your affirmation of me! Please find my response below and find my revisions in the resubmitted document.
Regarding the limitations of my research, I believe it is very necessary to make it more complete. I have rewritten my conclusion section, and its limitations are reflected in the diversity of user needs, information processing capabilities, and course evaluation mechanisms, as detailed in lines 502-512 of this article.

Reviewer 2 Report

Review Report

This is an exciting paper. However, some significant points must be addressed before it is accepted for publication.

1. In the abstract, add a few lines; what type of application does the author want to discuss in the article? Revise it.

2. Add a few sentences highlighting the significance of the suggested methodology in the abstract.

3. The reviewer suggested that the author write at least one paragraph on the importance suggested work and how much the developed technique is better than the previous one. Revise in introduction section.

4. The reviewer suggested that the author before writing any abbreviation must be defined. For example, “ACF” in line 57 is not defined.

5. What is meant by “RIB” in line 60?

6. Give some more information about Figure 1. for better understanding.

7. Give some explanation of Figure 2. at the bottom of the figure.

8.  What is meant by ? In Figure 3.

9. There is a punctuation mistake in line 199. Check carefully.

10. The meaning of sentences in lines 255-256 is not clear. Rewrite the sentences with clear meaning.

11. The author adds the research question in the introduction section.

12. Data in Figure 4. Should be in English?

13. Give some more explanation of Figure 5. for more better understanding.

 14. Need to discuss a comprehensive, detailed conclusion.

 

Remarks: After revising the above comments article will be recommended to accept. (Encourage to submit revision).

After revising the above comments article will be recommended to accept. (Encourage to submit revision).

Author Response

Dear Reviewer,
Thank you very much for taking the time to review this manuscript. Thank you very much for all your opinions and suggestions! Please find my response item by item below and find my revisions in the resubmitted document, highlighted as my revisions.

1.I have added in the abstract. I hope to provide an application that can meet the personalized learning needs of users and consider the semantic information of learning resources.

2.Also supplemented in the abstract. This method addresses the issue of resource diversity, and it is necessary to understand the different contributions of video and text learning resources to meeting learning needs.

3.In the introduction section of the article, specifically lines 96 to 116, I introduced the specific work of this article, elaborated on the improvements of this experiment, and emphasized the importance of the work at the end.

4.I'm sorry for the inconvenience caused by the abbreviations I used. In my future work, I will pay more attention to these details. ACF stands for Attention Collaborative Filtering, which can be found on lines 71-74.

5.Sorry again, RIB represents the recommendation framework for modeling the sequence and impact of micro behavior, specifically on lines 76-77.

6.I have rewritten the introduction section of Figure 1 and added some content, hoping that these can better assist in understanding. Specifically, on lines 163-177.

7.I have supplemented the information in Figure 2 and adjusted the position for a better understanding. Specifically, on lines 180 to 200.

8.I'm sorry that I can't see this image here. I'm not sure how to make any modifications. If possible, I hope to obtain a new image and give me a chance to make revisions to make the article more perfect.

9.This is my mistake, I have modified the punctuation marks. Especially on line 256.

10.I rewrote this sentence in the hope that its meaning would be clearer. Especially on lines 317-319.

11.Adding research questions is very necessary as it can make my article more complete. I added research questions in the introduction section. Especially on lines 40-48.

12.You are right, the data should be in English. I made changes to the data in the image and changed the color of the image to enhance its appearance.

13.I provided more explanation for Figure 5. Specifically, on lines 450 to 473.

14.I have rewritten the conclusion section, hoping it can be comprehensive and detailed. Specifically, on lines 482-512.

Reviewer 3 Report

This paper proposes a learning objection recommendation method based on combining KG and CF. In specifically, they first use TransD to get embedding of each entity on the KG, and derives the similarity between these entities. Then they also get the similarity between entities by using Item CF, a traditional CF algorithm. Finally, they establish the recommendation by assigning a normalized weight on these two similarities. However, this paper still has several flaws to be fixed, and here are my concerns:

1) The most concern is about the novelty. TransD is a typical KG embedding model which published in ACL 2015, and Item-CF is a much older method. Can the authors assert that combining these two methods will get a state-of-the-art performance? For instance, can authors make a comparison between the proposed method with KGAT [1], which is also a KG-embedding based CF model? Also, I cannot find a comprehensive performance comparison with other learning object recommendation models

2) Learning object recommendation is an important recommendation scenario which has been focused for several years. However, the reference is too shallow. Here I give some typical references the authors should read: [2-5]

3) There are two same title on subsections. Section 4.1 and section 4.2 are all “Experimental dataset”. Also, it is suggested to perform experiments on public datasets, such MOOCube.

 

 

[1] KGAT: Knowledge Graph Attention Network for Recommendation, KDD 2019

[2] Learning Behavior-oriented Knowledge Tracing . The 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD'2023)

[3] One Person, One Model—Learning Compound Router for Sequential Recommendation . The 22nd IEEE International Conference on Data Mining (ICDM), 2022.

[4] Personalized Employee Tranining Course Recommendation with Career Development Awareness . In Proceedings of the 27th World Wide Web Conference (WWW’2020)

[5] A hybrid e-learning recommendation approach based on learners’ influence propagation. IEEE TKDE, 2019

The language can be improved. There is Chinese character in line 127, and there is Chinese words Fig.4 without translation. Also, the reference format is not correct neither.

Author Response

Dear Reviewer,
Thank you very much for taking the time to review this manuscript. Thank you very much for all your opinions and suggestions! Please find my response item by item below and find my revisions in the resubmitted document, highlighted as my revisions.

1.I fully understand your concerns. In section 4.4 of the article, I added a comparison of algorithms, including the more advanced KGAT, CKE, and the benchmark algorithms TransD and ItemCF in this article. The results showed a slight improvement, as shown in Table 4. Subsequently, I analyzed a certain reason, which may be due to the greater need for specific interaction data between students and resources in the current experimental scenario. In this case, this article also used an implicit feedback model. Taking into account these reasons, we obtained a slight advantage.Specifically, on lines 388 to 434.

2.Thank you very much for providing me with these literature. I have read them one by one and cited them separately in the article. I believe the articles you provided have been very helpful to me and have deepened my understanding of learning recommendations. The following are the citation positions of typical references in the article:
[2] In lines 77-82, reference [12]
[3] In lines 42-44, reference [3]
[4] In lines 54-57, reference [5]
[5] In lines 60-62, reference [7]

3.I apologize for my negligence. I have changed the title of section 4.2 to 'Evaluating indicator'.Regarding the dataset, thank you very much for proposing such a professional dataset. I have already downloaded the dataset. However, due to time constraints, I did not apply this dataset this time. On the other hand, I would like to explore the impact of different learning resources on students in this article. I have not yet found a dataset with both video and text resource information. But I believe this is a direction for my future efforts, and I have added this to the limitations of the conclusion.

In summary, I am very grateful for your suggestions, which have been of great help to me both in terms of literature and methodology.

I have made modifications to the Chinese language on line 127, and after the modifications, it will appear on line 169.

Figure 4 has changed the language to English, and the references have standardized the style.

I apologize for the poor language in my manuscript. This time, I have made improvements in language and readability, hoping for a substantial improvement in language quality.

Reviewer 4 Report

Item "2. Related Work" should be renamed, since the sources have already been reviewed earlier.

You should double-check the names of the vectors in “there are corresponding projection vectors:?，??，?，??，?，??。”.

In Figure 1 it is not clear what values should be on the axes.

Paragraph 2.2 ends with formulas; it is better to add an explanation after them.

In Figure 2 there are some extra circles.

In Equation (3), the designation of the last "+" sign at the end is not clear.

The diagonal matrix shown in Table 1 is trivial. It can be omitted without loss of meaning.

Figure 4 is in a language other than English. Color coding is not clear.

Equations (9)-(11), which describe the calculation of F1-score, are well known. Their indication is unnecessary.

In Tables 4 and 5, the values for ? and K should be placed in the column heading.

There is no explanation for the bend in the blue graph in Figure 5a.

The quality of the text should be improved, including the use of capital letters in headings.

References are given in different styles.

Author Response

Dear Reviewer,
Thank you very much for taking the time to review this manuscript. Thank you very much for all your opinions and suggestions! Please find my response item by item below and find my revisions in the resubmitted document, highlighted as my revisions.

1.I will rename this section as "Methods" and focus on introducing the techniques used in this article.

2.I deeply apologize for the inconvenience caused by my mistake. I have rewritten this section. Specifically, on lines 163 to 177.

3.This figure refers to the original literature of the model, and the coordinate system mainly represents the vector space, so the values on the axis do not need to be clearly defined.

4.I have added some content after the formula, hoping to have a better understanding. Specifically, on lines 163 to 171.

5.I am very sorry for my carelessness. I have made modifications to Figure 2.

6.In lines 247-248 of the article, I added an explanation for the "+" symbol.

7.Thank you for your suggestion. I have deleted Table 1.

8.The data should be in English. I made changes to the data in the image and changed the color of the image to enhance its appearance.

9.I have deleted equations (9) - (11)

10.Your suggestion is necessary to make the table more reasonable. I have made modifications to Tables 4 and 5.

11.I have analyzed the cause of bending in Figure 5 (a), specifically in lines 462-473.

 

I have made improvements in language and readability, and the references have also been standardized in style.

Round 2

Reviewer 3 Report

I have to admit that the authors have addressed my concerns. 

Therefore, I only have one minor concern which I believe the authors should figure out before publish their work:

- I insist that Learning object recommendation is a long-term study and there are many paper published in famous venues. In the last round I only suggest 5 most representative and typical, I suggest the authors should give a more comprehensive literature review on this subject. Currently, threre are only 23 references in the revised version, but as we can observe there are usually more than 40+ refs in the top-tier LO recommendation papers. So the authors may make a better literatuer review. I can put some other usful suggestions but please note there are more important refs should be reviewed:

[1] Zhu, Y., Lu, H., Qiu, P., Shi, K., Chambua, J., & Niu, Z. (2020). Heterogeneous teaching evaluation network based offline course recommendation with graph learning and tensor factorization. Neurocomputing, 415, 84-95.

[2] Wenzheng Feng, Jie Tang, Tracy Xiao Liu, Shuhuai Zhang, and Jian Guan. Understanding Dropouts in MOOCs. AAAI-19.

[3]Jifan Yu, Gan Luo, Tong Xiao, Qingyang Zhong, Yuquan Wang, wenzheng feng, Junyi Luo, Chenyu Wang, Lei Hou, Juanzi Li, Zhiyuan Liu and Jie Tang. MOOCCube: A Large-scale Data Repository for NLP Applications in MOOCs. ACL-20.

[4] Zhu, Y., Lin, Q., Lu, H., Shi, K., Liu, D., Chambua, J., ... & Niu, Z. (2021). Recommending learning objects through attentive heterogeneous graph convolution and operation-aware neural network. IEEE Transactions on Knowledge and Data Engineering. 

Author Response

Thank you very much for recognizing my revised version. Learning object recommendation is a topic worth in-depth research. I have conducted a more comprehensive literature review. Thank you for recommending these valuable academic papers to me again. I have increased the number of references to 46, including the 4 you recommended to me and some other important references.
The details are as follows:
Regarding what you recommended to me:
[1] In lines 65-68, reference [19].
[2] In lines 32-33, reference [8].
[3] In lines 523-525, reference [46].
[4] In lines 68-72, reference [20].

Other important references:
Reference [1] [2] is on lines 24-25.
Reference [4] is on lines 28 to 29.
The references [6] [7] are in lines 31-32.
The reference [9] is in lines 35-38.
Reference [10] is on lines 39 to 40.
Reference [11] [12] is on lines 40-41.
Reference [14] is in lines 50-52.
Reference [23] is in lines 78-79.
References [27] [28] are in lines 96 to 98.
Reference [32] is in lines 107-108.
References [33] [34] are in lines 123 to 125.
The reference [35] is in lines 134-135.
The references [36] [37] are in lines 143-144.