[MMSEM] Image Annotation on the SW review

Dear all, with respect to 

ACTION: Ovidio, Stamatia to review http://www.w3.org/2001/sw/BestPractices/MM/image_annotation.html before mid October teleconf [recorded in http://www.w3.org/2006/09/14-mmsem-minutes.html#action02]

you can find below my comments.

The goals of the Image Annotation on the SW document are: i) to illustrate that deploying sw for image annotation is advantageous, and ii) to provide quidelines for SW-based image annotation (by means of proposed solutions and identified open issues), through the presentation of representative use cases and corresponding possible solutions.

Thus, under this context, a first remark would be that the added value brought by using sw technologies should be highlighted more, actually by better linking the content of the document itself. For example, in the introduction, the principal issues related to image annotation (and actually resource annotation in general) are very accurately presented, and then the notion of explicit and machine prossesable semantics introduced by sw is considered. Of course, it's quite obvious that explicit semantics entail certain advantages both in all types (content, media..) of annotations, but it'd be nice to show what the sw functionalities specifically add. The BigImage example given in owl is nice, but including for instance an example where owl is used to represent the semantics of an image decomposition into segments, some depicts property between segments and concepts, and, image and concepts, and having as available metadata that certain segments depict parts of a human body, then it is shown that sw enables to infer that this specific image depicts a person. this example might be too complitacated, but the idea is to be careful not just show that sw can be used to annotate images, which can be done using any xml  or whatever based standard or vocabulary, but that there is some added value. 

The selection of the use cases is indeed very good; they cover differing requirements in annotation. I'm not that convinced that the four categories classification is the most appropriate, as there are several, not orthogonal, dimensions along which this categorization could take place, such as professional versus personal, the type of annotations (global, region-based, low-level, descriptive, media-related and so on), the topics depicted, etc.

Section 3 briefly mentions vocabularies for image annotation, and more specifically mpeg7 and VRA. Since the discrimination among the different types of vocabulaties needed for image annotation has been clearly established earlier in the document, this section should either include some brief overview of the different categories' vocabularies (e.g., the structural ontologies by acemedia and minswap could be too such vocabularies, tv-antyime, etc.) or named differently. I would suggest categorizing them depending on the functionality covered, and mention the most representative ones pointing for further deatils on the Vocabularies Overview document. And as a general comment, care should be taken on how the terms vocabulary, ontology and metadata standard are used throughout the document. Teh last paragraph on the need sometimes present in image annotation to refer to specific parts rather than the whole image, should be moved imho in the image annotation issues section (1.1). 

Section 4, should be analogous to section 3 but for annotation tools. Not all aspects of existing tools functionalities need to be considered, and not in great detail either. 

With repsect to the example solutions, my main comment is that a uniform description/presentation should be followed. 

Some typos (complementing George's):

- TOC: 5..2 Use Case: large-scale.. -> capitilize the "l"
- Introduction: "technologies for image annotations" -> "technologies for image annotation"
- Image Annotation Issues: "The reader should be warned, however, that large scale, industrial strength image annotation is" -> "Large scale, industrial strength image annotation is however"
- Image Annotation Issues: "When lacking a crystal ball" -> "Consequently" ( i certainly applause the lyricism but maybe better to keep it more formal :-) )
- Image Annotation Issues: "which can vary wildly" -> "which can vary widely/significantly"
- Semantic Web Basics: "This section briefly describe" - > "This section briefly describes"
- Semantic Web Basics: "A complementary approach" ->
- Semantic Web Basics: "xmlns:rdfs="http://www.w3.org/2000/01/rdf-schema##"> remove redudant #
- Use case: Management of Personal Digital Photo Collections: "becomes quickly unused" -> "becomes quickly unmanageable"
- Vocabularies for Image Annotation: "to use for annotating image" -> "to use for image annotation"
- Available Tools for Semantic Image Annotation, Format of metadata: "are better appropriate" - > "are better suited"/"are more appropriate"
- URI and ID conventions (at 5.2 use case): "but when a the browser" -> "but when the browser"
- Text fields and control vocabularies: "The annotations below use three different uses" -> "The annotations below exeplify three different uses"

kind regards, 

Received on Wednesday, 11 October 2006 16:36:40 UTC