Deep Learning for Depression Detection from Textual Data
, Maha Abdelhaq 3, Raed Alsaqour 4,*, Sharnil Pandya 5 and Mueen Uddin 6Round 1
Reviewer 1 Report
In this paper, the authors proposed a productive model by implementing Long-Short Term Memory (LSTM) model to predict depression from the text. The proposed deep learning method is used to train textual data to identify depression from text, semantics, and written content. The authors should consider the following issues in their manuscript:
- The abstract section is inconsistent and does not reflect the main contributions of the manuscript. The authors should rewrite the abstract section to mention the main purpose of the paper, primary contributions, experimental results, and global implications.
- The paper needs intensive proofreading as it contains many long, inconsistent sentences. Besides, the manuscript contains many grammar errors and typos.
- It is highly recommended to remove the subsections in the introduction section. The contributions and the paper structure should be represented at the end of the introduction section without presenting them in individual subsections.
- The contributions are generic and do not introduce any novelty. The authors should rewrite the contributions precisely to show what is new in their research topic.
- The limitations of the current related work should be transferred to the end of the literature review section. Then, the authors should discuss how they overcame these limitations in their proposed system.
- The paper contains many undefined abbreviations, which can be considered a source of ambiguity for the reader. The authors should define any used abbreviation with its full text in its first place in the context. Then, the authors should use only the abbreviation throughout the rest of the paper.
- The discussed studies in the literature review section should be indicated by using the authors’ names, such as write “Smys and Raj [9] suggested ……” instead of “Authors in [9] suggested”
- The authors should discuss the used methodology in choosing the listed studies in Table 1. Besides, the listed studies should be referred to by the authors’ names in the first column.
- It is highly recommended to list all the used symbols in a table with their descriptions to follow the proposed methodology easily.
- The dataset is unbalanced and may affect the results significantly. The authors should solve the problem of the unbalanced dataset.
- It is not important to show the confusion matrics as the authors calculated accuracy, recall, precision, and F-measure.
- The authors should compare their proposed system with other systems from the current related work.
- At the end of the conclusion section, the authors should discuss their future research directions.
Author Response
Reviewer 1
Reviewer#1, Concern #1: The abstract section is inconsistent and does not reflect the main contributions of the manuscript. The authors should rewrite the abstract section to mention the main purpose of the paper, primary contributions, experimental results, and global implications.
Author response: Many thanks for noticing this issue. We can confirm that we have addressed this comment.
Author action: We have updated the abstract section as suggested (Section: Abstract, Page:1).
Reviewer#1, Concern #2: The paper needs intensive proofreading as it contains many long, inconsistent sentences. Besides, the manuscript contains many grammar errors and typos.
Author response: Many thanks for noticing this issue. We can confirm that we have addressed this comment.
Author action: We proofread the complete paper and removed mistakes.
Reviewer#1, Concern #3: It is highly recommended to remove the subsections in the introduction section. The contributions and the paper structure should be represented at the end of the introduction section without presenting them in individual subsections.
Author response: Thank you. We can confirm that this comment has now been addressed.
Author action: We have removed subsections from the introduction section and placed them as suggested. ( Page:2, line 49-65).
Reviewer#1, Concern #4: The contributions are generic and do not introduce any novelty. The authors should rewrite the contributions precisely to show what is new in their research topic.
Author response: Many thanks for pointing out this. We can confirm that we have addressed this comment.
Author action: We have made changes as suggested.( Page:2, line 49)
Reviewer#1, Concern #5: The limitations of the current related work should be transferred to the end of the literature review section. Then, the authors should discuss how they overcame these limitations in their proposed system.
Author response: Many thanks for pointing out this. We can confirm that we have addressed this comment.
Author action: We have made changes as suggested.(Page:33, line 90-105)
Reviewer#1, Concern #6: The paper contains many undefined abbreviations, which can be considered a source of ambiguity for the reader. The authors should define any used abbreviation with its full text in its first place in the context. Then, the authors should use only the abbreviation throughout the rest of the paper.
Author response: Many thanks for noticing this issue. We can confirm that we have addressed this comment.
Author action: We have updated the changes related to abbreviations.
Reviewer#1, Concern #7: The discussed studies in the literature review section should be indicated by using the authors’ names, such as write “Smys and Raj [9] suggested ……” instead of “Authors in [9] suggested”.
Author response: Many thanks for noticing this issue. We can confirm that we have addressed this comment.
Author action: We have modified the literature review as suggested. ( Page 2 and 3)
Reviewer#1, Concern#8: The authors should discuss the used methodology in choosing the listed studies in Table 1. Besides, the listed studies should be referred to by the authors’ names in the first column.
Author response: Many thanks for noticing this issue. We can confirm that we have addressed this comment.
Author action: We have made changes as suggested. ( Page 2 and 3)
Reviewer#1, Concern#9: It is highly recommended to list all the used symbols in a table with their descriptions to follow the proposed methodology easily.
Author response: Many thanks for noticing this issue. We can confirm that we have addressed this comment.
Author action: We have elaborated the symbols as suggested.
Reviewer#1, Concern #10: The dataset is unbalanced and may affect the results significantly. The authors should solve the problem of the unbalanced dataset.
Author response: Many thanks for noticing this issue. We can confirm that we have addressed this comment.
Author action: We have modified the suggested changes .
Reviewer#1, Concern #11: It is not important to show the confusion matrics as the authors calculated accuracy, recall, precision, and F-measure.
Author response: Many thanks for noticing this issue. We can confirm that we have addressed this comment.
Author action: We have modified the suggested changes.
Reviewer#1, Concern #12: The authors should compare their proposed system with other systems from the current related work.
Author response: Many thanks for noticing this issue. We can confirm that we have addressed this comment.
Author action: We have modified the suggested changes. (Page:9)
Reviewer#1, Concern #13: At the end of the conclusion section, the authors should discuss their future research directions.
Author response: Many thanks for noticing this issue. We can confirm that we have addressed this comment.
Author action: We have modified the conclusion section. (Page:11)
Reviewer 2 Report
The authors proposed an LSTM-RNN model to detect depression from text (in the case from Tweets). It can be helpful to detect mental disorders and suicidal tendencies before it's too late!
The authors did a great job by TABLE 1, summarizing different state-of-the-art methods. However, there are so many others that can be mentioned in the literature review (not necessarily in Table 1, but in the text) including:
Orabi, Ahmed Husseini, Prasadith Buddhitha, Mahmoud Husseini Orabi, and Diana Inkpen. "Deep learning for depression detection of twitter users." In Proceedings of the Fifth Workshop on Computational Linguistics and Clinical Psychology: From Keyboard to Clinic, pp. 88-97. 2018.
As an example, the above-mentioned study seems to be one of the most highly-cited ones in this field, about a hundred citation, but it has not been mentioned in the text.
Moreover, the reference [1] in the submitted manuscript, i.e.,
"Deep learning for prediction of depressive symptoms in a large textual dataset"
where its first author is in common with the submitted manuscript, has also proposed an LSTM- RNN approach for prediction of depressive symptoms in a large textual dataset. However, it is not described enough in the submitted manuscript. The authors should clearly state the differences between the their own proposed method and that of [1].
In the page 2, lines 40-43 are exactly repeated again in lines 50-52. Their next two sentences are also very similar.
Figure 1 needs more description in the text.
The use of English language is fine, however, it is recommended to be checked once again. For example, at lines 8,126 and 154, comma before the word "which" is missed. (a nonrestrictive clause in the middle or at the end of a sentence)
Author Response
Reviewer 2
Reviewer#2, Concern #1: The authors did a great job by TABLE 1, summarizing different state-of-the-art methods. However, there are so many others that can be mentioned in the literature review (not necessarily in Table 1, but in the text) including:
Orabi, Ahmed Husseini, Prasadith Buddhitha, Mahmoud Husseini Orabi, and Diana Inkpen. "Deep learning for depression detection of twitter users." In Proceedings of the Fifth Workshop on Computational Linguistics and Clinical Psychology: From Keyboard to Clinic, pp. 88-97. 2018.
As an example, the above-mentioned study seems to be one of the most highly-cited ones in this field, about a hundred citation, but it has not been mentioned in the text.
Author response: Many thanks for mentioning this. We can confirm that we have addressed this comment.
Author action: We have cited the above mentioned study in the literature as suggested by the reviewer.(Page: 3, lines 79-80)
Reviewer#2, Concern #2: Moreover, the reference [1] in the submitted manuscript, i.e.,
"Deep learning for prediction of depressive symptoms in a large textual dataset"
where its first author is in common with the submitted manuscript, has also proposed an LSTM- RNN approach for prediction of depressive symptoms in a large textual dataset. However, it is not described enough in the submitted manuscript. The authors should clearly state the differences between their own proposed method and that of [1].
Author response: Many thanks for noticing this problem. We can confirm that we have addressed this comment.
Author action: We have stated the difference between our proposed work and the study mentioned above.(Page:9, line 163-168)
Reviewer#2, Concern #3: In page 2, lines 40-43 are exactly repeated again in lines 50-52. Their next two sentences are also very similar.
Author response: Many thanks for noticing this problem. We can confirm that we have addressed this comment.
Author action: We have updated the lines as suggested and removes the repeated sentences (Page:2, line 53)
Reviewer#2, Concern #4: Figure 1 needs more description in the text.
Author response: Many thanks for noticing this problem. We can confirm that we have addressed this comment.
Author action: We have added the description of figure 1 in the text as suggested.(Page:2)
Reviewer#2, Concern #5: The use of English language is fine, however, it is recommended to be checked once again. For example, at lines 8,126 and 154, comma before the word "which" is missed. (a nonrestrictive clause in the middle or at the end of a sentence)
Author response: Many thanks for noticing this problem. We can confirm that we have addressed this comment.
Author action: We have modified the lines 8, 126 and 154 as suggested( Page:2, 6,8) and checked the complete paper.
Round 2
Reviewer 1 Report
The authors satisfied most of my concerns. I do not have more comments for them.
