Use Learnable Knowledge Graph in Dialogue System for Visually Impaired Macro Navigation
Round 1
Reviewer 1 Report
Authors provide a system architecture for supporting visually impaired users with a system for asking and hearing directions, etc. The system uses a collection of off-the-shelf tools (eg from google), and builds a triplet database of possible phrase groupings. The system needs to be tested and results generated for the paper to be complete.
Author Response
Point 1: Authors provide a system architecture for supporting visually impaired users with a system for asking and hearing directions, etc. The system uses a collection of off-the-shelf tools (eg from google), and builds a triplet database of possible phrase groupings. The system needs to be tested and results generated for the paper to be complete. 

Response 1: We have added the complete tests and revised the title of the manuscript to be fit the actual works. Since our work focuses on destination suggestions rather than complete navigation work, the test method is to plan three scenarios with different reasoning levels to verify the results of the system.
Author Response File: 
Author Response.docx
Reviewer 2 Report
The English language has to be improved.
The term Macro navigation does not refer to navigation as such. It instead refers to natural language processing, reasoning, and interface process.
The provided experimental results are not statistically meaningful. Besides, they contradict many other observations, where blind and severely visually impaired people are more confident and used to computer-human interfaces than sighted (blindfolded) people while using ETA systems for navigation.
This work does not explain how to use it in real-life navigation applications while guiding people through obstacles, different routes, indoor/outdoor destinations. Thus, it does not deal with navigation issues. Instead, it helps through dialogue to find out the target destination.
The authors should explain to the readers more thoroughly why a proactive ETA (electronic traveling aid) system's dialogue is needed instead of the blind user's natural speech interface control when he/she straightforwardly tells the system what he/she wants, and the system makes suggestions. For instance, the blind person says, 'I want to visit a Chinese restaurant,' and the ETA system gives titles of the nearby Chinese restaurants. In this way, the user is proactive and gives directions to the human-computer interface.
Author Response
Point 1: The term Macro navigation does not refer to navigation as such. It instead refers to natural language processing, reasoning, and interface process.
Response 1: "Macro navigation" may have different meanings in different environments. We use this term as a representation of walking navigation based on the below paper. We also learned from that paper and revised our article that added a dash between two words as "macro-navigation".
- Nicholas A.B.; Mark D.D. (University of Strathclyde, Glasgow, Scotland, G1 1XH, UK). Investigating context-aware clues to assist navigation for visually impaired people. Tel: +44(0)1415524400, Fax: +44 (0)1415525330 E-mail: {Nick.Bradley, Mark.Dunlop}@cis.strath.ac.uk, 2002.
Point 2: The provided experimental results are not statistically meaningful. Besides, they contradict many other observations, where blind and severely visually impaired people are more confident and used to computer-human interfaces than sighted (blindfolded) people while using ETA systems for navigation.
Response 2: Yes, we agree that severely visually impaired people are more confident and used to ETA which they familiar with, but because our tester is not familiar with the system we developed, he acted very cautiously. However, we have deleted these descriptions in the latest manuscript, because the focus of this paper is to get the destination recommendations, rather than the use of complete navigation for visually impaired people.
Point 3: This work does not explain how to use it in real-life navigation applications while guiding people through obstacles, different routes, indoor/outdoor destinations. Thus, it does not deal with navigation issues. Instead, it helps through dialogue to find out the target destination.
Response 3: Thank you very much for your comments, we also found this problem, so we have revised the title and content, emphasizing that it helps through dialogue to find out the target destination. But we have done a navigation system for blind people, which belongs to another project.
Point 4: The authors should explain to the readers more thoroughly why a proactive ETA (electronic traveling aid) system's dialogue is needed instead of the blind user's natural speech interface control when he/she straightforwardly tells the system what he/she wants, and the system makes suggestions. For instance, the blind person says, 'I want to visit a Chinese restaurant,' and the ETA system gives titles of the nearby Chinese restaurants. In this way, the user is proactive and gives directions to the human-computer interface.
Response 4: We have revised part of the introduction and made many descriptions in the experiment and conclusions to express the reasons and benefits of using the destination recommendations. Looking forward to your advice.
Author Response File: Author Response.docx
Reviewer 3 Report
The authors of this paper highlight the need of the visually impaired for macro navigation that can handle long-distance information to help them plan the journey except for nearby environment detection while walking. I appreciate that the authors conducted the current research in this important domain of developing macro walking navigation, which can be integrated with micro navigation to assist the visually impaired to arrive at the target safely. This is an interesting topic based on the learnable knowledge graph in the multi-turn dialogue system and the widely used GPS and GIS. Specifically, the authors use the knowledge graph as the basis of inference in the dialogue system and then update the knowledge graph so that the system gradually conforms to the user's background.
I find the study very valuable and scientifically documented due to its fullness, clarity and consequence in the flow of the paper. The manuscript is also well written and properly structured. The aim of this study and the research hypotheses are clearly stated in the Introduction section. The “Methods” and “Results” sections are well written.
Although the aim of this study and the research hypotheses are clearly stated in the Introduction section, the Discussion section should be more extended. Furthermore, the state-of-the-art is not analyzed in sufficient detail. This paper would receive help from several improvements:
- Specifically, in the “Discussion” section, the authors could link more thoroughly and successfully their findings with the results of previous researches. Although the authors mention in the “Discussion” section, that the general purposes dialogue system is still difficult to meet the needs and the concrete and feasible navigation dialogue system architecture which is proposed by them is very useful in practice, the authors could describe in more detail which gaps are covered with their research.
- Additionally, the literature review could be strengthened. While the authors did a nice job in the description of the methods and syntax analysis, they should introduce and describe some prior work in the field of the Macro navigation system.
- The “Conclusion” is missing, although the “Discussion” is too small. Even though “Conclusion” is not obligatory, due to the small length of the Discussion, it would be better to be included.
Kindly redo the paper following the above-mentioned suggestions to improve the paper and then resubmit it.
Author Response
Point 1: Specifically, in the “Discussion” section, the authors could link more thoroughly and successfully their findings with the results of previous researches. Although the authors mention in the “Discussion” section, that the general purposes dialogue system is still difficult to meet the needs and the concrete and feasible navigation dialogue system architecture which is proposed by them is very useful in practice, the authors could describe in more detail which gaps are covered with their research.
Response 1: Thank you for your suggestion. We have made a lot of searching about the recent research on macro-navigation, but found that the results are not fruitful. Therefore, we have integrated these findings into the discussion.
Point 2: Additionally, the literature review could be strengthened. While the authors did a nice job in the description of the methods and syntax analysis, they should introduce and describe some prior work in the field of the Macro navigation system
Response 2: We have added the literature on macro-navigation and ETA to the Introduction, hoping that it will be helpful for readers to understand the previous works.
Point 3: The “Conclusion” is missing, although the “Discussion” is too small. Even though “Conclusion” is not obligatory, due to the small length of the Discussion, it would be better to be included.
Response 3: We did have too few contents in the discussion and conclusion at the end. We have already made a lot of descriptions on the review of the paper and the improvement in the future. If there is any deficiency, please advise us.
Author Response File: Author Response.docx
Round 2
Reviewer 2 Report
The authors should provide more reasonable explanations to the journal reader:
1. What methodological value-added has this paper?
2. What new scientific insights, not published before, have been found?
3. Why meaningful empirical testing statistics is not provided?
Author Response
Point 1: What methodological value-added has this paper?
Response 1: For visually impaired navigation, we design a knowledge graph for use in natural language human-computer dialogue system. This system can effectively improve the accuracy of NL-based human-computer interaction.
Point 2: What new scientific insights, not published before, have been found?
Response 2: When visually impaired people are walking outdoors, they need more guidance and assistance. Natural language is the most ideal interface for human-computer interaction, but the inherent problems of natural language are uncertainty and ambiguity. For these problems, using knowledge graphs is a good method.
Point 3: Why meaningful empirical testing statistics is not provided?
Response 3: This research designed a NL dialogue system for visually impaired navigation with a knowledge graph, and also implemented this system. In the initial stage, we invited the visually impaired people to verify the usability of this system. After passing the IRB application in the future, we will devote more time and resources to real-environment testing to obtain more measured data. Finally, we hope to make improvements for the blind’s outdoor mobility.
Author Response File: 
Author Response.pdf
Reviewer 3 Report
Although the authors have made many changes to the manuscript, I have made the following notifications,
- The changes have been made very quickly, especially in the sections of the discussion and conclusion, which could be analyzed and described more thoroughly. Even the title of the section conclusions is misspelt. As a result, these sections are not well written and need to be rewritten.
- Additionally, the literature review should be strengthened. There are papers in this research field that explain the terminology explicitly and they are not included in the references. As a consequence, macro-navigation is not well explained in the whole document, how it is used in real-life navigation applications and how this is useful for indoor and outdoor destinations. There is no good connection between these terms. Finally, the research is more oriented to natural language processing, reasoning and the navigation dialogue system.
Author Response
Point 1: The changes have been made very quickly, especially in the sections of the discussion and conclusion, which could be analysed and described more thoroughly. Even the title of the section conclusions is misspelt. As a result, these sections are not well written and need to be rewritten.
Response 1: Thank you for your reminder so that we can fix this obvious title error. We have rewritten the discussion and conclusion for a more complete analysis and description. We have also reviewed the findings of this paper in more depth, so that readers can understand the contribution and limitations of this paper. If there is anything that should be improved, I hope to get your advice again.
Point 2: Additionally, the literature review should be strengthened. There are papers in this research field that explain the terminology explicitly and they are not included in the references. As a consequence, macro-navigation is not well explained in the whole document, how it is used in real-life navigation applications and how this is useful for indoor and outdoor destinations. There is no good connection between these terms. Finally, the research is more oriented to natural language processing, reasoning and the navigation dialogue system.
Response 2: Based on your suggestion, we have added a literature reference about macro navigation, the reference number is [3], which is an earlier and complete description of macro navigation that we can find. In order to explain the application of macro navigation in more detail, we added a real outdoor test in the Result section and summarized a brief discussion for the test. We hope that readers can better understand the macro navigation through these experiments. Thanks again for your comments, let us know if there is any content need to be modified.
Author Response File: Author Response.docx
Round 3
Reviewer 2 Report
.
Reviewer 3 Report
The changes are acceptable. First of all, the subsection which is entitled «3.2 Outdoor Test» in the"Results" section is being added. Discussion and Conclusion have been also improved.
Round 4
Reviewer 2 Report
Now the article more or less fits the journal scope.
