Applying Linked Open Data for Green Introduction

Tracking #: 478-1673

Takahiro Kawamura
Takahiro Kawamura (2)

Responsible editor: 
Guest editors Semantic Web Interfaces

Submission type: 
Full Paper
The growing environmental consciousness has been causing the concern for urban agriculture and greening business. However, plant cultivation in an urban restricted space is not necessarily a simple matter, and it may extinct depending on species. On the other hand, if it overgrows, it could lead to break the vegetation balance of the surrounding environment. Therefore, we propose an Android application called Green-Thumb Camera, which queries the plants to fit environmental conditions from the LOD cloud based on smartphones sensor information, and then overlays its form in the space using AR to show an image of the mature plant for amateurs. In this paper, after description of LOD content generation method and the application details, we show the evaluation of accuracy of LOD content and usability of the application.
Full PDF Version: 

Major Revision

Solicited Reviews:
Click to Expand/Collapse
Review #1
By Martin Voigt submitted on 24/Jun/2013
Major Revision
Review Comment:

Basically, the paper proposes a new and interesting use case for using Linked Open Data (LOD) in combination with smart-phones: the identification of suitable plants to cultivate them in the current environment. Therefore, the authors describe how they created the required LOD content and how the use it. The article closes with an evaluation of the content generation and of the usefulness of the entire application.

However, the article is a more practical one with a low theoretical originality, thus I tended to re-categorize it as application report. The authors should clearly state what the scientific contributions of the article are. Furthermore, some questions arose and I identified problems with regard to the writing and the quality of some figures, which are both detailed per section in the following. All in all, the work requires a major revision to be part of the special issue of the semantic web journal.

## Abstract ##
- “after description of LOD” -- “after the description of LOD”

## Introduction ##
- “greenery” is used to much in the first column
- “… expertise were available … “ -> “is available”
- What is “3DCG”? It is mentioned nowhere.
- Acronyms should be introduced _after_ the complete word(s), thus, switch the order for “AR (Augmented Reality)” and “LOD (Linked Open Data)”. The problem occurs in other sections as well. Please check it.
- The challenges and the (scientific) contribution of the article are not discussed!

## Proposal of Green-Thumb Camera ##
- Page 2: Introductions between headlines of 2. and 2.1, 2.2 and 2.2.1 are missing.
- P. 2: Although problems to solve are discussed, the technical ones are not clearly stated in 2.1.
- P.2: Fig. 1 is hard to read
- Names of applications, classes … should be emphasized at least on their first occurrence.
- P. 4: Fig. 2 should be moved to page 2 or 3. Furthermore, the arrows cross some labels, which decrease their readability.
- P. 2: “10,000+” -- more than 10.000
- P. 3: Fig. 3, here the graphics in the middle, are not readable.
- P. 3: The enumeration of the properties “{ … }” at page 3 is hard to read and maybe not necessary.
- There is always a missing space before a reference, not only in Sect. 2.
- P. 3: “LOD generation is as follows” -- “works as follows”
- P. 3: Why is Google used for ranking? Why 100 pages?
- P. 3: Where do you get the synonyms for the property names within the bootstrapping method? An example would increase the understandability.
- P. 3: The extraction of triple from unstructured text using “dependency parsing” seems to be easy. In my opinion, this approach neglects many special cases, e.g., occurrence of multiple plant or property names in a sentence, sentiments … Is there a solution for this cases?
- P. 3: Remove spaces in < plantname, property, value > and the following triples.
- P. 3: I did not understand the part “Furthermore, for exclusion of … and the second-best.”
- P. 3: I did not understand the last two sentences as well.
- P. 4: Fig. 4, the font size is a bit too small as well.
- P. 4: Remove the after the footnote behind “for Android^2,”
- P. 4: “To begin with we believe…” is not to understand
- P. 4: The usage of the “Sunlight” factor is really questionable since it may hardly depend on the weather or the day time. I think, there is a database required which holds an average sunlight value for a specific region.
- P. 5: The factor “Planting season” comprises an high uncertainty as it is shifted using one or two months. Is it crucial?

## Evaluation ##
- P. 5: Intro is missing between section and subsection.
- P. 6: Again, the clustering, which is introduced at page 3, is hard / not to understand.
- P. 7: The plants are hard to see in Fig. 6.
- P. 7: The description of the lines is missing in Fig. 7.

## Related Work
- P. 8: I did not understand the first method to generate LOD. Maybe example papers would assist to understand the mentioned approach.
- P. 8: “AGROVOC does not include the knowledge for plant cultivation” -- Please, add some information about the concepts of the vocabulary and how you distinguish your approach. Why was it not possible to reuse the vocabulary?

Review #2
By Tru Cao submitted on 28/Jun/2013
Review Comment:

The paper presents an application on smartphones that allows a user to visually check what plants match the user's surrounding environmental conditions. A user's smartphone is equipped with sensors to obtain surrounding conditions that are to be matched to Linked Open Data (LOD) about plants, which previously generated.

As such, there are two concerned problems. The first one is generation of LOD about plants and their related information. The second one is matching surrounding conditions to the corresponding ones in the generated LOD to recommend suitable plants. However, the presentation about proposed solutions to these problems are quite verbose, not showing any sophisticated technicality.

More fundamentally, given instant conditions surrounding a place, it is not sufficient to tell if a particular plant is suitable to that place. Rather, it depends on changing conditions within a certain period.

For the presentation, there are following drawbacks:
- Grammatical mistakes: e.g., in the abstract, "may extinct" => "may be extinct", or "lead to break" => "lead to breaking".
- Abbreviations should not be used in an abstract (e.g., AR, LOD) without their full forms.
- Fig. 3 should be placed after Fig. 2.
- In Section 4, citation should not be used like "presented by T. Mitchell at AAAI10, ...".

Lastly, the paper's contents have little things to do with Semantic Web.

Therefore, although this application could be useful for users without gardening expertise, the paper does not have notable research/technical contributions.

Review #3
Anonymous submitted on 05/Jul/2013
Major Revision
Review Comment:

This paper describes an interesting application of LOD and particularly the semi-automtic extraction of LOD and how that can be used to provide a service to end-users.
I have a few suggestions/queries regarding the paper:

- The writing at various parts in the paper is hard to follow or does not read well. A full edit of the text would be required before the paper could be published
- Given the LOD context set up at the beginning of the paper I was very surprised by the small number of plants in the sample. It begs the question what would be the implication in terms of data accuracy and user experience if a more realistic sample was used.
- it is not clear where the ground truth comes from for the data in table 1. Was an expert or source text used to establish the ground truth?
- It is not clear how light levels are used in the system given that light levels obviously vary with weather and time of day. This does not seem to be taken into consideration
- It is not clear how the data in first study was collected. What questions/observations were used to come up with the ratings.
- It is not clear what happened in the second study, What were the five participants doing for 30 hours? How was this interpreted to reach the conclusions?