Deep Interest Context Network for Click-Through Rate

Round 1

Reviewer 1 Report

Overall, the paper is very exciting and its content is absolutely relevant.

The background is explained acceptably (more details would be better). Overall result is of great interest and very exciting. 

However, both the Results and Conclusions are not covered in enough detail.

You read the paper looking forward to the results and a well-founded discussion of the results and then you think that everything is described too briefly.

Also more information on the approach would be nice to see.

Author Response

Point 1: The background is explained acceptably (more details would be better).

Response 1: We sincerely thank you for this point of view. We read more literature and then elaborate more and more detailed background and history of the development of the field of recommender systems in the introduction section. (Change in lines 24-27)

Point 2: However, both the Results and Conclusions are not covered in enough detail. You read the paper looking forward to the results and a well-founded discussion of the results and then you think that everything is described too briefly.

Response 2: For this issue, we are very grateful for your valuable comments. We modified the presentation of the experimental results and created a bar graph to express the superiority of the model more clearly and visually (Change in lines 362-365). At the same time, in the conclusion section, we add more detailed conclusions derived from the experimental results, and refer to the literature to elaborate the future development direction of the industry to verify that the model in this paper is fully in line with the mainstream research direction. Finally, the outstanding problems and shortcomings of the model are given, and the future work is given an outlook (Change in lines 399-407). 

Reviewer 2 Report

Deep Interest Context Network (DICN) works based on two interesting factors. Does your system can select these attributes dynamically

Also is it possible to change the input attributes from two to more?? If so then what may be the effect. 

Try to add answer to these questions in Introduction or conclusion section

Author Response

Point 1: Does your system can select these attributes dynamically?

Response 1: Thank you very much for your question. After experiments our system can dynamically select properties, for which we have reworked lines 83-84 and 141-143 of the article to make additions.

Point 2: Also is it possible to change the input attributes from two to more? If so then what may be the effect. Try to add answer to these questions in Introduction or conclusion section.

Response 2: We sincerely thank you for your constructive questions and comments on the paper model. For this point, we consider for the time being only two environmental attributes, season and weekday, while other attributes are not considered but theoretically possible, although there is a risk of overfitting. We put the answer to this question in the final conclusion section in the outlook of future work, adding other environmental attributes to be studied later if necessary. At the same time, the outlook of the development direction of the industry is added to verify that the model fully fits the key development direction of the industry. (Change in lines 391-392 and 405-407)