Deep Collaborative Recommendation Algorithm Based on Attention Mechanism

Round 1

Reviewer 1 Report

The authors suggested an attention-based deep collaborative recommendation system. First, before feeding the user-item representations into the DNNs, they used the attention mechanism to apply varying weights to the user-item representations, capturing the hidden information in implicit feedback. The user-item representations with accompanying weights were then fed into the representation learning and matching function learning modules. Finally, they concatenated the learned prediction vectors from various dimensions to forecast the matching scores. Here are some of my thoughts:

1-The organization of the introduction part is weak and uneven. Combine paragraphs to get four strong paragraphs. This section also needs further citations. The following key citations should be used by the authors:

https://ieeexplore.ieee.org/abstract/document/9262856

https://www.sciencedirect.com/science/article/abs/pii/S2210670722004061

https://link.springer.com/article/10.1007/s00521-022-07424-w

https://www.sciencedirect.com/science/article/abs/pii/S092523122100477X

https://www.mdpi.com/2072-4292/12/7/1149

2-The sentences have some sort of misunderstanding. The authors employ abbreviations in various areas of the paper, although they do not reflect the stated terms. for example linear embedding as interaction function (DMF), a deep collaborative recommendation algorithm based on attention mechanism (DACR)... and so on.

3-You should not leave any sections blank. Section 2 is an example. Fill it with appropriate sentences. Do not jump forward to another section.

4-If you use formulas from other publications, you must appropriately reference them.

5-The authors should go into further detail about their method and findings. This version is inadequate.

6-The work's implication is absent from the conclusion section.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

I am wondering that how the proposed model differs from the state-of-the-art models. The authors mentioned that they dont consider auxilliary information, this I think limits the model capability because in real world, the metadata information matters a lot.

A few references can be cited and explain, why auxilliary information is needed or not for the proposed architecture by authors.

https://dl.acm.org/doi/10.1145/3178876.3186175

I think main figure 2 needs more explaination, esp the order of layers and the input output sequence,

The selection of dataset Movielens is too naive,

I would ask authors to please mention the key takeaway from the results of negative sampling that what is the main lesson learnt by using negative samples.

I think, experiments be performed on another dataset also and then compared.

 

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

It can be accepted.

Reviewer 2 Report

authors have addressed most comments, though improvements in experiments can be seen in future works.