The Zaragoza’s Knowledge Graph: Open Data to Harness the City Knowledge

Round 1

Reviewer 1 Report

Thanks for revising the manuscript according to my comments.

Now the paper looks much better, but still some issues need to be improved:

1) Figure 1 needs more descriptions, e.g., a summary to briefly introduce each component. Otherwise, it is hard to clearly understand the whole workflow. Usually, we extract the knowledge from different data sources to generate a knowledge graph and then build applications on top of it. Here, in figure 1, it seems that besides rdf generation from some datasets, you still have other data sources to query. What's the relationship between the previous generated rdf data and other data sources?

2) The sentence in the response letter: "The knowledge acquisition is manually configured in the data management system and all data are mapped according to vocabularies" should be clearly put in the paper.

3) I cannot understand the sentence "Then, when a new data request is received the knowledge extraction is automatically executed by the programing layer and the Knowledge Graph generated" in the response letter. Do you mean your knowledge graph are built in a dynamic way? However, you still show the statistics of a static knowledge graph in table 3.  Writing should be carefully improved. Still a lot of parts in the paper make me confused.

Author Response

Dear reviewer,

Thanks for your valuable feedback. We have now addressed your comments, so that this submission considers most of them, as described in more detail in the attachment.

Kind regards.

Author Response File: Author Response.pdf

Reviewer 2 Report

Adequate improvements have been made to the paper. Extensive comments are addressing some of my initial concerns. The contribution is valuable enough to be published and its presentation was significantly improved.

Author Response

Dear reviewer,

Thanks for your final comments.

Kind regards.

Round 2

Reviewer 1 Report

Thanks for revising. Only one comment:

In the second paragraph of Sec. 4, "Figure 3" should be "Figure 1".

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.

Round 1

Reviewer 1 Report

This paper introduces the Zaragoza’s Knowledge Graph, including the generation of the KG and some applications of the KG. I think it is a good application paper which uses knowledge graph technologies to organize the government data in a semantic way. It is good to see unifying, reusing and exploiting the city data in this project. It is a good contribution to the research community of open data or linked open data.

This work has several unclear parts which need to be improved. My comments are as follows:

1) For the section of Knowledge Graph generation, it is not clear that are there any automatic techniques applied in knowledge extraction and knowledge fusion from different data sources? In my understanding, the government should have a large-scale of relational data, textual data and others, so how to perform knowledge extraction with existing ontologies and knowledge fusion is a key point. Or authors should clearly clarify that this process is conducted manually.

2) In Figure 1, some icons' meanings are not clear.

3) What about the quality of the created knowledge graph?

4) For the presented applications, since I do not know much about the Spanish, it is hard for me to evaluate their quality. I suggest the website should at least provide the English interface. This could benefit foreign tourists or business men to use this website. 

5) Presentation should be improved.

Reviewer 2 Report

The paper presents interesting Action Research where Linked Data-driven services have been developed for the Zaragoza city council.

This is nice pragmatic work based on real world use cases and is definitely worth publishing.
However, some adjustments should be made to present it less like a journalistic report and more like a scientific report. I would prefer to see this framed as a Design Science project, structured according to Design Science research - from requirements to evaluation.

The figures focus too much on showing front-end screenshots (in non-English language) instead of depicting design decisions or supporting technical discussion. Ontologies should be presented with dedicated visualization tools (VOWL, Protege plugins etc.) or at least as a UML metamodel (class diagram depicting domains and ranges for the main properties). I understand that these are huge vocabularies (hundreds of properties, it seems) but at least some isolated fragment could be discussed - perhaps closely related to the front-end screenshots (examples of SPARQL queries relevant to the screenshot, examples of reasoning/axioms, examples of interoperability/reconciling challenges that were met, considering how many datasets were involved).

This paper sometimes reads like it advertises a product - instead, it should focus on technical details, ideally summarized in diagrammatic form. Figure 1 is the only attempt at providing something like this, but it's very vague - what API are we talking about? what are the things on the right side? what do the arrows represent - order? data exchanges? queries? This should be done with standard means (e.g. a UML deployment or component diagram for architecture, a separate BPMN or activity diagram for the workflow/generation pipeline)

Some information is superfluous - Table 2 is just a list of well known formats, available in almost every SPARQL endpoint. Don't waste paper space with such tables, a simple enumeration would suffice.

There's no dedicated section for Evaluation. The benefits of the proposal are mentioned throughout the paper but it would help to have them summarized in a dedicated subsection.

Related work could be extended with the large body of work that exists now on Knowledge Graph development. Some examples:
- https://link.springer.com/chapter/10.1007/978-3-030-29551-6_51
- the KG series of proceedings: https://www.springer.com/gp/book/9789811331459
- http://koreascience.or.kr/article/JAKO201809538046051.page