Advancements in Complex Knowledge Graph Question Answering: A Survey

Round 1

Reviewer 1 Report

This is a survey of research regarding complex knowledge graph question answering (C-KGQA) regarding the various strands of research methods: graph metric-based, graph neural networks-based, as well as those based on Large Language Models (referred as Pre-trained Language Models PLM in the manuscript). The authors also discuss the theoretical preliminaries regarding C-KGQA research, resources and evaluation, as well as emerging trends. The research analysis overall is of good quality and novelty, and the specific comments below relate to areas where I think the authors could improve on.

 

1. There are too many spelling and grammar errors, and in particular inconsistent use of capital letter terms or acronyms (e.g. sometimes C-KGQA and sometimes KGQA), to the degree of affecting the readability of the article. Some acronyms such as PLM should be expanded the first time they show up. I strongly suggest the authors to substantially revise the article text and correct these issues.

 

2. Categorization of research trends. The first two strands (metric-based and graph neural network) are classic strands which is OK on its own, however the cited works are also classic and not sufficiently recent, which limits the novelty of the authors research. The strand of pre-trained language models isn't particularly novel either, as this is late 2023 where many of the cited works have been around for sometime. In general I suggest adding survey on more recent works especially new ones in the classic strands but with Large Language Model-based elements.

 

3. On resources and evaluations. Whilst the discussion regarding the robustness of C-KGQA models is sensible, there is just one cited work. What are some of the research attempting to address these raised issues?

 

4. Survey from the aspect of real-world applications of knowledge graph question answering is completely lacking. Having a dedicated section on this would complement the current manuscript that are overly focused on the theoretical and methodological aspects.

 

5. The conclusion section. Instead of saying "It also implicitly indicates the bottleneck challenge in this field", the authors should appropriately and directly discuss the limits and challenges regarding the generalizability of a C-KGQA model. In addition, regarding the emerging trends identified by the authors, are there any pilot / preliminary works related to those trends?

 

I have mentioned the issues regarding the writing and use of English in the previous section.

Author Response

On behalf of all the contributing authors, I would like to express our sincere appreciations of your letter and reviewers' constructive comments concerning our manuscript entitled "Advancements in Complex Knowledge Graph Question Answering: A Survey".

 

Author Response File: Author Response.pdf

Reviewer 2 Report

This paper presents a literature review for the topic of complex knowledge graph question and answering. The authors categorized previous works into three categories and discussed their pros and cons in detail. The authors also summarized the knowledge graph, dataset, and evaluation metrics used in previous works and talked about the trends for the future work. Overall, the paper structured well, but there are a few comments need to be addressed:

1. The authors provided a relatively comprehensive background for the topic of knowledge graph QA in the sections of “Introduction” and  “Preliminary”, and then started to talk about existing methods. However, the novelty and the major contributions of the paper are not clear. Is there any existing literature review and how this paper differs from existing works?

2. The "Nobel prize" example discussed in the "Introduction" section is not described clear enough. What is the "previous triple" and "another triple" refer to?

3. For Table 2, it is more informative if authors could include a column summarizing pros and cons of each method. I also suggest adding reference for each method in the table.

4. I suggest adding similar tables as Table 2 for other two types of methods for comparison purposes.

5. The authors discussed the trends for this research field, however, I did not see clear challenges for the existing works. I suggest authors include a section to discuss the challenges.

There are a few grammatical errors and formatting issues need to be addressed:

1. There are a few abbreviations need to be defined in the main paper. For example:

"KGQA" in line 31;

"GM" methods in line 67;

"PLM" in line 120.

2. There are sentences starting with lowercase letter or starting without a space after period. For example:

Line 47 ". according to..." should be ". According to...";

Line 153 "[29].The success" should be "[29]. The success".

3. Line 82-84 are not completed sentences.

 

Author Response

sincere appreciations of your letter and reviewers' constructive comments concerning our manuscript titled "Advancements in Complex Knowledge Graph Question Answering: A Survey".

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

The authors have made revision changes that are satisfactory to my earlier comments. I think this manuscript is a good scientific work now.

Reviewer 2 Report

I think the authors addressed all my comments. I appreciate their efforts.