Review Reports

Round 1

Reviewer 1 Report

The paper proposes an interesting solution based on a new index called Lattice-tree in order to improve performances of Spatial Concept Queries SCQ. SCQ were defined previously in literature and the authors claims that their method can overcome the limitation of previous solutions since they use the lattice structure to represent many-to-many relationships between objects and attributes. 

The idea is nice, but the presentation and the formalization of the problem contains many imprecisions. In particular:

- It is not clear why you suppose that that spatial objects should be ordered and which order is applied? (Lines 216-217)

- It is not clear which regions are represented in the lattice-tree. In the example presented in the introduction the lattice structure shown in figure 2 refers to a specific region of interest. How these regions are stored in the index?

- What is the meaning of the parameter delta? I really do not understand the definition of this parameter and its motivation. (Lines 256-259)

- Which is the criterion that is applied for grouping the spatial objects in the same node? How do you store regions in the lattice-tree (Algorithm 1).

- If you index is just a R-tree with additional information in the nodes describing the relationships among objects and attributes by means of a lattice structure, then just present it in this way!

- Which is the meaning of step 2 in figure 4? What is the meaning of “satisfying a lattice node”? How can the point p be in the lattice? It is the query point? You mean that you generate a new lattice starting from the point p and the lattice that you have in the nodes? How do you compute such lattice?

- Which is the cost of traversing the tree nodes?

- In section 4.2 the choice of parameters delta and r need more motivations.

- How are the numbers presented in Table 2 correlated to the numbers reported at lines 248-349? How should I read table 2? If level 1 is the level of the leaf nodes, what is the meaning of the numbers: 2, 4, 65293? It is absolutely not clear!

- The fact that parameters delta and r need to be set by considering the dataset at hand is a drawback of the proposed solution; at least you should discuss the cost that is requested to tune these parameters every time you change the considered dataset.

- Can the proposed index be updated?

Author Response

Dear reviewer,

Thank you for your recognition of the importance of the work.

Please see the attachment.

Authors

Author Response File: Author Response.pdf

Reviewer 2 Report

 

The authors have presented an innovative approach to indexing spatial data. The results clearly show the problems with the spatial textual Big Data. Overall, the paper is very well written, both structurally and linguistically. The introductory chapter provides the reader with enough information about previous relevant research and its implications for the present study. The proposed workflow is easy to follow, and the results obtained are clearly presented. The experimental design is sufficiently described to be repeatable. The conclusion is supported by the results obtained.

I have some specific suggestions on how the manuscript could be improved:

Please correct the abstract according to the instructions for authors. Add in the last sentence what you found and why your work is important.

It would be much more meaningful if the figures were in color

You should mention the programming languages and libraries you used e.g. line 323.

The conclusions need to be expanded to interpret your results.

Finally, I would encourage you to publish the code and data you used for the experiment and analysis as part of the article as I think this will be of great use to other researchers.

Author Response

Dear reviewer, 

Thank you for your encouraging words and recognition of the importance of the work. 

Please see the attachment.

Authors

Author Response File: Author Response.pdf

Reviewer 3 Report

The authors present a novel hybrid index structure and an algorithm that can be used in the spatial keyword query of textual big data. The paper has following major issues:

1. The proposed index structure was not sufficiently compared with the existing approaches.
2. The conclusion is based on only one data. It is unclear how the proposed structure works on other data.
3. As the level increases, the number of tree objects increase exponentially. The practicality of the approach in general is unclear.
4. The English needs a careful revision, avoiding spelling errors and inconsistencies in the styles.

Author Response

Dear reviewer, 

Thank you for your recognition of the importance of the work.

Please see the attachment.

Authors

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

The authors have addressed all the issues that I have highlighted in the first review.

However, there are still two weak points that must be addressed in order to accept the paper for publication:

1) English language and style must be improved.

2) The presentation of the formalism in section 3.2 must be improved.

In particular for point 2 my suggestion is to rewrite the section as follows (I use the latex syntax for the formal expressions):

(i) describe the reference sets: \cal{K} representing the set of possible features (that are called attributes in the paper) that a location can have or not. The term "attribute" make the reader think that the attribute can assume different values, while in the paper the attribute represents a feature (for example, "IS OPEN") that can be true or false. I suggest to introduce also a set \cal{P} containing all the locations p.

(ii) define the set \cal{D} as follows: \cal{D} = \{ d_i = <p,K> | p \in \cal{P} and K \subset \cal{K} \}. Add an example of d_i, like d_i=< p_i, {supermarket, is_open} >

(iii) define a Spatial Textual Concept (STC), by saying that a Spatial Textual Concept c is a tuple <r,D,K,I>, where

• r is the considered spatial region,
• D \subset \cal{D}, is as set of spatial objects contained in r
• K \subset \cal{K} represent the common attributes of the spatial object of D
• I \in D × K: represents the pairs (d,k) indicating that the spatial object d has the attribute k

(iv) define the operators f and g (avoiding the term "equation" that is not correct here). These operators are used here to specify a constraint: that is, if we compute the f(D) if the STC is correct then f(D) must be equal to K. Thus the definition of f and g should be rewritten as follows:

• f(D) = {k \in \cal{K} | \forall d \in D, (d,k) \in I },  
• g(K) = {d \in \cal{D} | \forall k \in K, (d,k) \in I }

In order to be correct an STC c=<r,D,K,I> must satisfy the following constraint: f(c.D) = c.K and g(c.K) = c.D

(v) in the definition of the concept lattice you should start with a pair of STC! This part should be rewritten as follows:

• given two STC c_1=<r,D_1,K_1,I_1>  and c_2=<r,D_2,K_2,I_2> covering the same region r, the following partial order relation is defined:

c_1 ≤ c_2 ⇔ D_1 ⊆ D_2 ( ⇔ K_2 ⊆  K_1 )

• based on this the Concept Lattice L considering the set C of STCs existing in the spatial region r, can be formally defined as follows:

L = (r,C,≤)

where ≤ is the partial order relation defined above.

• the authors could also add a comment saying that: "for sake of readability in Figure 2 each STC <r,D,K,I> reported in the nodes of the lattice is represented only by listing the set of object D and the common attributes K".

Author Response

Please see the attachment.

Author Response File: Author Response.pdf