MKBQA: Question Answering over Knowledge Graph Based on Semantic Analysis and Priority Marking Method

Round 1

Reviewer 1 Report

Summary of the paper:

The authors proposed a knowledge graph based question answering method (mainly in computer science domain) that used WordNet for semantic query expansion along with priority marking algorithm.

Strong points:

1.     The motivation of this paper is good.

2.     The paper is well organized and easy to read.

3.    In Section 4, the experiments are interesting. I enjoyed this section, as it provides some impressive results. Experiments conducted on a reasonable sized dataset which showed the effectiveness of the proposed method. A minor suggestion is that, the authors can mention the default values of some parameters (if applicable) that are used in the experiments. 

Specific comments:

1.     Please briefly explain the algorithm propsed in Section 3 with it's complexity.

2.     Please mention the tools you used for data preprocessing with proper references.

3.     It will be good if you expand the explanation of "Question understaing error" (Table 12). 

4.     Please add a Conclusion section. 

Author Response

Thank you again for all the valuable comments and suggestions, which help us improve the quality of our manuscript and give us more insights into the research. We have responded to your suggestion in detail in the attachment (response to reviewer1.pdf). We hope these revisions rise to your expectations.

Author Response File: Author Response.pdf

Reviewer 2 Report

Dear authors,

Thank you for submitting your paper to Applied Sciences Journal. I believe your paper is an informative paper that can be published after a major revision.

Comments:

1.     The introduction  does not adequately describe the main problem in the literature, and how the paper has tackled it? Please expand the introduction,

2.     An adequate literature review and a clear gap identification have been tried to be conducted. However, authors have ignored some research which has been done in the area. I strongly recommend the authors to provide a more comprehensive literature review in the introduction section, particularly the definition of domain-specific KG, etc. The following paper is recommended:

·         “Domain-specific knowledge graphs: A survey", Journal of Network and Computer Applications (JNCA), Vol. 185, Issue: 1, https://doi.org/10.1016/j.jnca.2021.103076,

 

 

3.     Figure 3 – step 3, there is a small box titled, step 3 as well ? Please clarify

4.     Why SimCSE is used rather than other methods? Also give a reference to it

5.     A thorough editorial check and English improvement are needed. Please kindly proofread the entire manuscript.

 

6.     The conclusion part is missing; which questions are answered, what is the value/originality/contribution of the paper, how the proposed method answers the research questions that previous methods are not able to answer? 

Author Response

Thank you again for all the valuable comments and suggestions, which help us improve the quality of our manuscript and give us more insights into the research. We have responded to your suggestion in detail in the attachment. We hope these revisions rise to your expectations.

Author Response File: Author Response.pdf

Reviewer 3 Report

1. Paper Summary: This paper presents a question & answer model using knowledge graph. The authors first used logic rule to convert natural language intro triples. Then, WordNet is used to expand the semantic meaning of entities in the discovered triple. Finally SimCSE-based method is used when there are conflicts in the query. The proposed methods is called MKBQA. Experimental results show that the MKBQA out performs collected baseline models.

2. Strength: The methods proposed in this paper are clearly described. The detailed math formulas are given, and the implementation details on different datasets with different parameter thresholds are described. The English style is easy to understand in this paper. The comparison between the proposed model and baseline models are fair and convincing.

3. Weakness: Why the authors use red text several times in the paper? To indicate the importance of the corresponding statements? This is not necessary and strange, which should be corrected. Also, the abstract does not mention any experimental results. Now that the proposed MKBQA model out-performs the baseline models, the authors should mention that in the abstract.

4. This is basically a fair paper applying knowledge graph to question and answer. Although there are other deep learning approaches which can be applied to larger benchmarking databases, the proposed model still out-performs the baseline models on the selected dataset (CSQA & QALD3). Hence, after the revision on the above defects, I still recommend to accept this paper.

 

Author Response

Thank you again for all the valuable comments and suggestions, which help us improve the quality of our manuscript and give us more insights into the research. We have responded to your suggestion in detail in the attachment. We hope these revisions rise to your expectations.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Authors have addressed my comments, 

Author Response

Thank you again for all the valuable comments and suggestions, which help us improve the quality of our manuscript and give us more insights into the research. In the future, we will take our research work more seriously and strive for greater breakthroughs.