Towards the Semantic Enrichment of Trajectories Using Spatial Data Infrastructures
, Valéria Cesário Times 1, Jugurta Lisboa-Filho 3 and Chiara Renso 4Round 1
Reviewer 1 Report
Dear colleagues,
thank you for this highly valuable contribution "towards the semantic enrichment of trajectories using spatial data infrastructures". For my point of view this is a highly relevant work for the semantic extension of our data and its enabling of data integration.
I do not have substantial input. The paper is well structured. The pragmatic examples are given in the end. At some point it would have been helpful to have some pragmatic examples in the text in order to support the reading. At least a reference to the example of manual semantic enrichment could be helpful.
The example of a user with smartwatch that is given in the text (introduction, approach and example) allows to grab the topic of the contribution in a glance. Suddently it is clear what semantic enrichment means. But isn´t the selection of datasets by their metadata too rough? Aren´t there more information useful to select a dataset for the semantic enrichment and for a data integration? What are the detailed requirements in terms of semantic enrichment?
What is the role of ontology in your process of semantic enrichment?
In figure1 you describe the architectural model. Could you highlight any requirements concerning the interfaces, like the role of RDF for metadata?
When using the track of the user, how does the model deal with measurement errors? Is there any method you could propose?
All the processes and algorithms given, request some kind of validation.
Is the set of fields for the STMO document enough for the semantic data integration? Do you have any observations on the requirements and the behaviour?
Are any changes needed in existing standards, that you have used? Are the standards still appropriate? I assume that this novel approach could come up with needed adoptions.
Author Response
Response to Reviewer 1 Comments
Point 1: Thank you for this highly valuable contribution "towards the semantic enrichment of trajectories using spatial data infrastructures". For my point of view this is a highly relevant work for the semantic extension of our data and it enabling of data integration.
Response 1: Thanks. The authors are grateful for the comment.
Point 2: I do not have substantial input. The paper is well structured. The pragmatic examples are given in the end. At some point it would have been helpful to have some pragmatic examples in the text in order to support the reading. At least a reference to the example of manual semantic enrichment could be helpful. 

Response 2: Done. Thanks for the suggestion. A bibliographic reference for an example of manual semantic enrichment has been included (see line 708). We have also included a new paragraph that discusses the content of the aforementioned reference (see lines 707 – 711).
Point 3: The example of a user with smartwatch that is given in the text (introduction, approach and example) allows to grab the topic of the contribution in a glance. Suddently it is clear what semantic enrichment means. But isn´t the selection of datasets by their metadata too rough? Aren´t there more information useful to select a dataset for the semantic enrichment and for a data integration? What are the detailed requirements in terms of semantic enrichment?

Response 3: Initially, we would like to emphasize that semantic data integration is not part of the scope of our work. Recently, several Open Geospatial Consortium (OGC) initiatives have been developed to propose new interoperable and reusable standards on the Web through the OGC API. Thus, the use of these new standards together with microservices can facilitate dataset selection from their metadata, as they tend to improve the interoperability and integration of SDI's data and services and use API Gateway to manage application requests.
In addition, other information can be useful for selecting a dataset, such as information about trajectory segments and means of transport. For semantic enrichment, metadata can describe detailed information (i.e., a descriptive summary about the data) about the dataset, contain information about Points of Interest (POI), Regions of Interest (ROI), Lines of Interest (LOI), Activities, opening hours of POIs, stop points and movement, Dataset bounding rectangle, trajectory category, algorithms and methods used in semantic enrichment, and thresholds used in algorithms and methods.
Point 4: What is the role of ontology in your process of semantic enrichment?

Response 4: In this research, the concepts of supervised learning and classification algorithms were used. The use of ontology is not part of the scope of this work, but it has been included as future work to allow the retrieval and integration of semantic trajectory data (See lines 996 – 997). The ontology can be used to integrate semantic trajectory data, improving decision-making in semantic enrichment applications.
Point 5: In figure1 you describe the architectural model. Could you highlight any requirements concerning the interfaces, like the role of RDF for metadata?
Response 5: In this work, we do not use specifications with RDF schemas. The use of ontologies and RDF schemes for trajectories demand previously established definitions to represent the trajectory domain. An ontology that serves a certain user may not be useful for another user who performs the same trajectory, as both can perform activities and have different movement objectives. The use of ontologies and RDF is highlighted in the related works of this manuscript as research topics in SDIs to provide data retrieval and integration solutions (See lines 253 – 257). Figure 1 represents the model of the first SDIs. In this Figure, we can highlight the role of RDF in metadata interfaces and web services. RDF can be used for integration of different metadata standards or conversion between metadata standards and offers the possibility to define the semantics of your metadata. Web services can be described by RDF schemas. Therefore, RDF aims to ensure interoperability in SDI environments and the available interfaces.
Point 6: When using the track of the user, how does the model deal with measurement errors? Is there any method you could propose?
Response 6: In this work, error handling methods were not considered. We know that this research topic is widely discussed in the literature to solve GPS equipment measurement problems. However, the proposal of these methods is beyond the scope of our work. Recent studies on SDI indeed report the importance of documenting holistic trajectories. This type of trajectory requires real-time collection services, which can use SDI's semantic enrichment services, for example. At this point, we only visualized the possibility of using error handling methods, as the Internet of Things (IoT) equipment tends to produce many errors, demanding that these errors be handled during data collection. This could be an interesting future study.
Point 7: All the processes and algorithms given, request some kind of validation.
Response 7: Done. This point was included as a future research topic for process and algorithm validations (see lines 992 – 996). The authors intend to validate the STMO documentation processes in SDI and the algorithms proposed in this research. However, it is noteworthy that preliminary studies to validate the weighted average algorithms (algorithm 3 – line 6) and the central point (algorithm 3 – line 8) were previously discussed by the authors themselves in [51].
Point 8: Is the set of fields for the STMO document enough for the semantic data integration? Do you have any observations on the requirements and the behavior?
Response 8: In this work, we do not consider semantic data integration. Although the discussion on data integration is interesting for the semantic enrichment of trajectories, this subject is beyond the scope of this research. This question of yours sets precedents for new research topics.
Point 9: Are any changes needed in existing standards, that you have used? Are the standards still appropriate? I assume that this novel approach could come up with needed adoptions.
Response 9: Thanks for the discussion. The new ISO 19115:2014 norm (https://www.iso.org/standard/53798.html) is the standard used repeatedly in this manuscript. The ISO 19115:2014 standard is the latest update of the old ISO 19115:2003, but with a lot of flexibility in the use of its fields, aiming to facilitate the documentation of different datasets. This standard is appropriate and there are no newer updates. The new ISO 19115:2014 makes the use of its mandatory fields more flexible, which made it easier to create a minimal set of fields to document semantic trajectories. Therefore, the ISO 19115:2014 standard is relevant to our research. However, new investigations may be carried out in the future, as with the growth in the use of holistic trajectories, there is the possibility of collecting new types of semantic information, such as sensor data. This requires further analysis to decide which existing fields of the ISO 19115:2014 standard should be used or to propose the adoption of new fields. The idea is to have a flexible and universal standard for semantic trajectories, which can be proposed in future work on the validation of trajectory documentation processes.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
The authors have written a very interesting and high-quality article, which will be interesting and useful to many scientists. The article is presented correctly. Congratulations on the article.
Author Response
Response to Reviewer 2 Comments
Point 1: The authors have written a very interesting and high-quality article, which will be interesting and useful to many scientists. The article is presented correctly. Congratulations on the article.
Response 1: The authors are grateful for the comment.
Author Response File: Author Response.pdf
Reviewer 3 Report
The article's goal is to provide tools to manage STMOs with an SDI, and more precisely to carry out semantic enrichment and documentation of these trajectories. The author describe a global solution to adapt and complement traditional SDIs to enable the management of STMOs. Their solution relies on the use of Web and semantic micro-services, the adoption of a new concept of SDI, named SDI4Trajectory, that supports the semantic enrichment of trajectories, and the use of an easy-to-use metadata standard with a reduced number of fields, based on the ISO 19115-1: 2014 standard. The authors also describe algorithm to carry out semi-automatic or manual enrichment of trajectories.
The global structure of the article is very clear, and the authors give ample justifications for the interest of their proposal, and the technical choices they recommend.
The two main technical contributions, the semantic enrichment process and the documentation process, are described in detail. In particular, the authors provide specific algorithm to carry out the data enrichment process. Section 5 describes an experiment that evaluates the reliability of the semi-automatic semantic enrichment process (by comparison with manual annotations). The sample size for this experiment is small (15 tracks), which prevents the formulation of fine-grained conclusions beyond the fact that the semi-automatic enrichment process looks globally reliable, but the global argumentation of the artcile does not depend on this experiment, so the small sample size does not affect the global quality of the article. The authors give enough indications to make the reproduction of this experiment possible.
Below are more specific remarks:
- Definition 1 is unsufficiently clear to me. As it is written, it could lead the reader to understand that an action performed by a user is a micro-service. It should be rewritten to clearly state that the micro-service only allows the addition of manual semantic annotations.
- Figure 4 : step (1) of the process invoked several times, it would be useful to have an explicit unmbering of the steps on Figure 4 (as is done in Figure 7 for example)
- Definitions 3 and 4 describe functions (or properties) of API4Trajectory and SDI4Trajectory, and as they are written, the denomination « Definition » sounds improper. If authors want to keep this denomination, they should at least reformulate :
Definition 3 : API4Trajectory is an API that allows…
Definition 4 : SDI4Trajectory is a SDI that allows/encompasses…
- l. 451 : Computation of clusters : in this paragraph, it is stated that clusters are computed based on the speed of the user and a speed parameter given by the user, which could mean that a cluster is a set of points where speed does not exceed a pre-defined parameter (which would be reasonable definition). The subsequent discussion involving a similarity criterion and similarity values muddies this understanding, and suggests that a cluster needs to have points of uniform speed, which is not very intuitive.
- l. 466 : Semantic points : is there a specific reason to use semantic points and not a convex hull or a bounding rectangle ? It seems to be more straightforward to compute POIs and distances to POIs with semantic points. If so, the authors should give the reason for this choice as soon as the choice is made. Such a justification is all the more crucial because the authors recognize that the choice of POI and the final annotation are very sensitive to the definition of the semantic point.
- For Table 2, the definition for the Hits (%) column (the percentage of correct hits in carrying out the comparison) is too vague, a more precise definition (or a formula), should be given.
- l. 816-818 : In addition, these 18 attributes proposed for STMO documentation enable to automate the filling of 80% of the fields listed in Table 1.
I am not sure I understand this phrase properly. I understand that the fields listed in Table 1 are exactly the 18 attributes described above ; if that is the case, the formulation of this sentence is not clear enough, as the structure of the sentence suggests they could be different items. It would also be useful to state which fields can not be completed automatically.
English grammar and spelling is generally correct ; there are still a few grammar errors and missing words:
l. 14 : to validation → to validate
l. 41 : semantically enrichment → semantical enrichment
l. 203 : that difficult → that make the exchange of data among applications difficult
l. 308 : a word is missing : to which have paid little attention → to which studies/researchers have paid little attention
l. 382 : that is, the so-called holistic trajectories → that is, a so-called holistic trajectory
l. 402 : allow → allows
l. 441 : make them → makes them
l. 681 : see 9 → see Figure 9
l. 779 : three-stop points → three stop-points
Author Response
Response to Reviewer 3 Comments
Point 1: The article's goal is to provide tools to manage STMOs with an SDI, and more precisely to carry out semantic enrichment and documentation of these trajectories. The author describe a global solution to adapt and complement traditional SDIs to enable the management of STMOs. Their solution relies on the use of Web and semantic micro-services, the adoption of a new concept of SDI, named SDI4Trajectory, that supports the semantic enrichment of trajectories, and the use of an easy-to-use metadata standard with a reduced number of fields, based on the ISO 19115-1: 2014 standard. The authors also describe algorithm to carry out semi-automatic or manual enrichment of trajectories.
Response 1: Thanks. The authors are grateful for the comment.
Point 2: The global structure of the article is very clear, and the authors give ample justifications for the interest of their proposal, and the technical choices they recommend.
Response 2: Thanks. The authors are grateful for the comment.
Point 3: The two main technical contributions, the semantic enrichment process and the documentation process, are described in detail. In particular, the authors provide specific algorithm to carry out the data enrichment process. Section 5 describes an experiment that evaluates the reliability of the semi-automatic semantic enrichment process (by comparison with manual annotations). The sample size for this experiment is small (15 tracks), which prevents the formulation of fine-grained conclusions beyond the fact that the semi-automatic enrichment process looks globally reliable, but the global argumentation of the article does not depend on this experiment, so the small sample size does not affect the global quality of the article. The authors give enough indications to make the reproduction of this experiment possible.
Response 3: Done. The use of a consolidated database was included as future work (See lines 992 – 994). This database belongs to the Geolife project with almost 18 thousand trajectories.
Point 4: Definition 1 is insufficiently clear to me. As it is written, it could lead the reader to understand that an action performed by a user is a micro-service. It should be rewritten to clearly state that the micro-service only allows the addition of manual semantic annotations. 

Response 4: Done. The authors appreciate the suggestion. Definition 1 has been rewritten. (See lines 364 – 366).
Point 5: Figure 4: step (1) of the process invoked several times, it would be useful to have an explicit numbering of the steps on Figure 4 (as is done in Figure 7 for example).
Response 5: Done. The authors appreciate the suggestion. See Figure 4 on page 7. We also made the same change in Figure 2 – Page 4.
Point 6: Definitions 3 and 4 describe functions (or properties) of API4Trajectory and SDI4Trajectory, and as they are written, the denomination « Definition » sounds improper. If authors want to keep this denomination, they should at least reformulate:
Definition 3: API4Trajectory is an API that allows…
Definition 4: SDI4Trajectory is an SDI that allows/encompasses…
Response 6: Done. Suggestions accepted. Definitions 3 and 4 were rewritten to justify the adoption of the term “definition”. (See lines 405 – 408 and 414 – 416).
Point 7: l. 451: Computation of clusters: in this paragraph, it is stated that clusters are computed based on the speed of the user and a speed parameter given by the user, which could mean that a cluster is a set of points where speed does not exceed a pre-defined parameter (which would be reasonable definition). The subsequent discussion involving a similarity criterion and similarity values muddies this understanding, and suggests that a cluster needs to have points of uniform speed, which is not very intuitive.
Response 7: Done. The authors rewrote the paragraph to improve understanding of the use of speed in cluster formation (see lines 452 – 466). The velocity threshold criterion was explicitly considered in the definition of cluster calculations to avoid understanding that clusters need uniform velocity points to be built. Therefore, adjustments to the paragraph were made to improve the readability and clarity of the text.
Point 8: l. 466: Semantic points: is there a specific reason to use semantic points and not a convex hull or a bounding rectangle? It seems to be more straightforward to compute POIs and distances to POIs with semantic points. If so, the authors should give the reason for this choice as soon as the choice is made. Such a justification is all the more crucial because the authors recognize that the choice of POI and the final annotation are very sensitive to the definition of the semantic point.
Response 8: Done. The authors added a new paragraph (See lines 468 – 479) that presents the justifications for choosing the semantic point approach in our research. In time, the authors reinforce that the choice of this approach does not exclude others, as we consider that there are other ways (such as the use of the bounding rectangle) that are complementary to each other and can be used together to improve the inference of the semantic point.
Point 9: For Table 2, the definition for the Hits (%) column (the percentage of correct hits in carrying out the comparison) is too vague, a more precise definition (or a formula), should be given.
Response 9: Done. The authors agree that the paragraph was confusing. To solve this point, we inserted a formula and a new sentence to improve reader understanding (See lines 800 – 803).
Point 10: l. 816-818: In addition, these 18 attributes proposed for STMO documentation enable to automate the filling of 80% of the fields listed in Table 1. I am not sure I understand this phrase properly. I understand that the fields listed in Table 1 are exactly the 18 attributes described above; if that is the case, the formulation of this sentence is not clear enough, as the structure of the sentence suggests they could be different items. It would also be useful to state which fields cannot be completed automatically.
Response 10: Done. The authors rewrote the sentence indicated by the reviewer (see lines 839 – 842). In addition to this change, the authors added a new sentence to improve the readability of the paragraph (see lines 834 – 837).
Point 11: English grammar and spelling is generally correct; there are still a few grammar errors and missing words:
- 14: to validation → to validate
- 41: semantically enrichment → semantical enrichment
- 203: that difficult → that make the exchange of data among applications difficult
- 308: a word is missing: to which have paid little attention → to which studies/researchers have paid little attention
- 382: that is, the so-called holistic trajectories → that is, a so-called holistic trajectory
- 402: allow → allows
- 441: make them → makes them
- 681: see 9 → see Figure 9
- 779: three-stop points → three stop-points
Response 11: Done. The authors are grateful for the corrections. Please check lines 14, 41, 203, 308, 383, 403, 441, 694, and 797. In addition, the authors requested a new review by the English language professional.
Author Response File: Author Response.pdf
