VETS Visualization Gallery Metadata Proposal

Version: 0.2 draft
Date: 5/16/02
Author: Markus Stobbs, VETS
Phone: 303.497.1238

Please send comments and changes to mstobbs@ucar.edu.

Introduction

This document proposes a metadata approach that would effectively describe the image and animation assets of the Visualization and Enabling Technologies (VETS) section within the Scientific Computing Division (SCD) of the National Center for Atmospheric Research (NCAR). These visualizations are currently available in our Gallery at: http://www.scd.ucar.edu/vets/vg/.

The primary driver of this proposal is our plan to redesign the Gallery, storing metadata about visualizations in a database and providing search, browse, and streaming capabilities. Since we will be describing the properties of visualizations in order to catalog our digital media assets and provide a search capability, it seems wise to go a step further and align with XML and RDF standards to enable future interoperability of the Gallery with the Semantic Web envisioned by the WWW Consortium (W3C). The intention is also to contribute to the wider effort of the Web Advisory Group (WAG) and THREDDS to create metadata standards for UCAR, NCAR and atmospheric science.

Assumptions

Application Flow

The Visualization Gallery web application will have an application flow starting at a search page and leading to a specific visualization or a research project detail page. It will consist of the following page templates which will all be database-driven.

• Easy Search Page
• Advanced Search Page
• Search Results - a list of visualization hits grouped by research project
• Visualization Detail Page
• Research Project Page

Use RDF and Multiple XML Namespaces

This document assumes a basic knowledge of RDF and XML. For more background on these topics, see:
http://www.w3.org/TR/rdf-primer/

RDF, which is expressed in XML, has an advantage over straight XML in that it provides a rich capability to indicate entity relationships. Since the very nature of atmospheric data and visualizations involves multiple hierarchical and network relationships, RDF is most appropriate for our metadata description needs.

RDF makes it possible for diverse communities of experts to contribute to a web of machine-readable vocabularies. Each community creates a schema in the form of an XML namespace that describes their domain. Multiple XML namespaces can live happily side-by-side in the same file. It is assumed that wherever possible we will use existing XML namespaces which are already in widespread use or show promise as emerging standards. It is tempting to play God and invent an entire schema and vocabulary according to how you see the world, but if we are ever to achieve the dream of the Semantic Web, we will need to agree on some core namespaces and be good citizens about using them.

Start with Dublin Core

The purpose of Dublin Core is to support Internet-wide searching with a small set of commonly used fields. The concept is to limit the number of fields so that the probability of finding resources can be supported across many systems. Dublin Core already has widespread support, and is used liberally in the W3C documentation for XML and RDF. It is safe to assume that the major search engines will build in support for Dublin Core.

After first describing our content's basic properties with Dublin Core, we will provide a more robust description of the Geoscience-specific properties using metadata standards in our field. Where we don't find existing standards that meet our needs, we will either extend an existing standard or create our own namespace(s). Without a doubt, UCAR and NCAR are in a position to be leaders in the creation of namespaces for Atmospheric Science.

RDF describes resources. For this application, it is assumed we have 4 resources of the following Dublin Core types to describe:

1. Image, be it a graphic, animation or video
2. Interactive Resource, for Flash files and animations with interactive features (currently we have none)
3. Dataset (optional) which is the data source used to create the image
4. Collection (optional) which is the research project the image is a part of

Note that this data is relational. A single research project will contain multiple visualizations, and each visualization will consist of multiple file formats (QuickTime, Real Media, MPEG, JPEG thumbnail). The metadata presented here simplifies matters by presenting a flat-file view with the exception of the occasional tags: relation, subClassOf and subPropertyOf.

It is worth noting that the Dublin Core Working Group is currently working on standards for element and value qualifiers. Until they make their formal recommendations, there is some question about exactly how to refine DC elements and encode values. Many of the examples below use "dcterms" from the DCMI proposed recommendation Expressing Qualified Dublin Core in RDF/XML.

Use Controlled Vocabularies

Wherever possible, we will endeavor to select values from a controlled vocabulary in the interest of consistency and machine-readability. One example of this is choosing subject values from the Library of Congress Classification (LCC) rather than typing in a free-form subject.

Namespaces Used in this Recommendation

The following namespaces have proven useful for the purposes of this recommendation. I am currently researching science and multimedia namespaces to choose ones appropriate for our needs. These are covered in Appendix 3.

Search Metadata

By starting with the metadata which users are most likely to want to search by and then identifying the properties they will want to see displayed in a detail view, including entity relationships, we will focus on the most important metadata for our actual implementation.

The primary purpose of describing visualizations with metadata will be to enable users (and agents) to search for images and animations of interest at a granular level. Especially as the number of visualizations in the Gallery expands, granular searching will be necessary to retrieve relevant search results. This issue is even more vital in the case of agents which may be searching all Geoscience sites or the Internet as a whole.

Browsing through the existing Gallery, it seems the following properties and values would be useful for search purposes. In other words, these properties are good candidates for an advanced search page. In addition to those listed, a keyword field on the advanced search and easy search pages would search many of the title and description properties listed in the Display Metadata section that follows this one, as well as the subject properties below. XML elements left blank indicate properties for which a suitable namespace has not yet been found. "ucar" and "ucarterms" are imaginary namespaces that are used to indicate where ucar might fill a void by creating a namespace of its own to describe properties for which there is no good existing namespace for our needs.

Visualization (Image) Display Metadata

In addition to the important metadata above which enables granular searching, there is detail metadata that the user will want displayed after finding an image or animation of interest. For example, with animations, it is vital to know how the visualization duration maps to that of the model duration in order to understand how time has been compressed. A 3D animation which is 2 minutes in duration, with a standard rate of 30 frames per second, where each frame is a timeslice of 1 minute in the simulation model represents a model duration of 1 hour. In other words, 1 hour of model time has been compressed into 2 minutes of visualization time.

The search metadata above would be included in the display view as well. We may find that some of the display metadata deserves to be elevated to the status of search metadata.

We also need to capture the important relationships between the visualization and its related objects. The power of RDF is describing the relationship between entities in a way that enables the Semantic Web envisioned by the W3C to become a reality. This will be a web in which objects know how they relate to each other, so one can inquire about related objects to discover a broader context, drill down into more detail, or just digress to a related topic. For example:

• A visualization file is often just one in a set of related visualizations that belong to a research project. Seeing that visualization in the context of the research project or paper is vital to understanding its full meaning.
• A visualization is created from a dataset. Wouldn't it be great to be able to get to that dataset in one click rather than stage a Google search for it.

Research Project (Collection) Metadata

There is some repetition of properties below for the collection when compared with the image resource. This is not yet an exhaustive list of research project properties.

Embedded Metadata

One of the interesting developments in metadata thinking is to embed metadata within each object where file format allows for that. There are several benefits to this approach:

• If an object is downloaded, it retains it's metadata, even if re-distributed.
• Search engines and agents of the future will be able to crawl and acquire metadata for each file, enabling the Semantic Web.
• Metadata stored only in databases and not within the object itself can elude crawlers. This means the object would not be indexed in search engines.
• Metadata stored only in databases can become orphaned from the object when file directories are moved or content changes owners.

For performance reasons and in the interest of rapid development, it may be necessary to start with a database approach and move to a search engine approach as semantic-enabled search engines become available. However, by embedding metadata in objects now, we will be prepared for the future Semantic Web.

There are already standards for embedding metadata in JPEG, PNG, and other graphic formats. The Adobe XMP documentation includes a PDF on embedding metadata which provides in-depth information for developers on this topic. Embedded metadata standards for video...

Other Metadata

Finally, there are the properties that are unlikely to be of interest to the average or even technical user, but which are included in current metadata standards for truly robust object description. Where this metadata is generated automatically by software applications, it can be considered labor free and might as well be captured where automation makes it a zero cost scenario. This category is included here mainly as a container for properties whose real value needs to be discussed to decide whether they should be included in our implementation or not.

Appendix 1: Metadata Expressed in RDF

<?xml version="1.0" ?>
<smil xmlns = "http://www.w3.org/TR/.../SMIL-Boston.dtd">
 <head>
 <meta id="meta-smil1.0-a" name="Publisher" content="W3C" />
 <meta id="meta-smil1.0-b" name="Date" content="1999-10-12" />
 <meta id="meta-smil1.0-c" name="Rights" content="Copyright 1999 John Smith" />
  <metadata id="meta-rdf">
   <rdf:RDF
       xmlns:rdf = "http://www.w3.org/1999/02/22-rdf-syntax-ns#"
       xmlns:rdfs = "http://www.w3.org/TR/1999/PR-rdf-schema-19990303#"
       xmlns:dc = "http://purl.org/metadata/dublin_core#"
       xmlns:smilmetadata = "http://www.w3.org/AudioVideo/.../smil-ns#" >
<!-- Metadata about the SMIL presentation -->
   <rdf:Description about="http://www.foo.com/meta.smi"
       dc:Title="An Introduction to the Resource Description Framework"
       dc:Description="The Resource Description Framework (RDF)
		 enables the encoding, exchange and reuse of structured metadata"
       dc:Publisher="W3C"
       dc:Date="1999-10-12"
       dc:Rights="Copyright 1999 John Smith"
       dc:Format="text/smil" >                
       <dc:Creator>
          <rdf:Seq ID="CreatorsAlphabeticalBySurname">
             <rdf:li>Mary Andrew</rdf:li>
             <rdf:li>Jacky Crystal</rdf:li>
          </rdf:Seq>
       </dc:Creator>
       <smilmetadata:ListOfVideoUsed>
          <rdf:Seq ID="VideoAlphabeticalByFormatname">
              <rdf:li Resource="http://www.foo.com/videos/meta-1999.mpg"/>
              <rdf:li Resource="http://www.foo.com/videos/meta2-1999.mpg"/> 
          </rdf:Seq>
       </smilmetadata:ListOfVideoUsed>
       <smilmetadata:Access LevelAccessibilityGuidelines="AAA"/>
   </rdf:Description>
<!-- Metadata about the video -->
   <rdf:Description about="http://www.foo.com/videos/meta-1999.mpg"
        dc:Title="RDF part one"
        dc:Creator="John Smith"
        dc:Subject="Metadata,RDF"
        dc:Description="RDF basic fonctionalities"
        dc:Publisher="W3C Press Service"
        dc:Format="video/mpg"
        dc:Language="en"
        dc:Date="1999-10-12"
        smilmetadata:Duration="60 secs"
        smilmetadata:VideoCodec="MPEG2" >
        <smilmetadata:ContainsSequences>
           <rdf:Seq ID="ChronologicalSequences">
             <rdf:li Resource="http://www.foo.com/videos/meta-1999.mpg#scene1"/>
             <rdf:li Resource="http://www.foo.com/videos/meta-1999.mpg#scene2"/> 
           </rdf:Seq>
         </smilmetadata:ContainsSequences>
   </rdf:Description>
<!-- Metadata about a scene of the video -->
   <rdf:Description about="#scene1"
        dc:Title="RDF intro"
        dc:Description="Introduction to RDF fonctionalities"
        dc:Language="en"
        smilmetadata:Duration="30 secs"
        smilmetadata:Presenter="David Jones" >
        <smilmetadata:ContainsShots>
           <rdf:Seq ID="ChronologicalShots">
             <rdf:li>Panorama-shot</rdf:li>
             <rdf:li>Closeup-shot</rdf:li>
           </rdf:Seq>
        </smilmetadata:ContainsShots>
   </rdf:Description>
  </rdf:RDF>
 </metadata>
<!-- SMIL presentation -->
 <layout>
    <region id="a" top="5" />
 </layout>
 </head>
 <body>
 <seq>
   <video region="a" src="/videos/meta-1999.mpg" >
     <area id="scene1" begin="0" end ="30"/>
     <area id="scene2" begin="30" end ="60"/>
   </video>
   <video region="a" src="/videos/meta2-1999.mpg"/>
 </seq>
 </body>
</smil>


<rdf:RDF
  xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#"
  xmlns:dc="http://purl.org/dc/elements/1.1/"
  xmlns:dcterms="http://purl.org/metadata/dublin_core_qualifiers#"
  xmlns:mpeg7="http://www.mpeg7.org/2001/MPEG-7_Schema">
  <rdf:Description about="http://www.dstc.edu.au/videos/98-02-20.mpg">
       <dc:Title>SBS World News</DC:Title>
       <dc:Creator>Special Broadcasting Service</DC:Creator>
   	<dc:Subject>
       	<rdf:Bag>
				<rdf:li>machine-readable catalog record formats</rdf:li>
				<rdf:li>applications of computer file organization and
				 access methods</rdf:li>
      		</rdf:Bag>
   	</dc:Subject>
       <dc:Publisher>Special Broadcasting Service</DC:Publisher>
       <dc:Contributor.Presenter>Indira Naidoo</DC:Contributor.Presenter>
       <dc:Format DC:Scheme="IMT">video/mpg</DC:Format>
       <dc:Type>Image.Moving.TV.News</DC:Type>
       <dc:Language>en</DC:Language>
       <dc:Date>12/05/98</DC:Date>
       <dc:Format.Length>30 mins</DC:Format.Length>
       <mpeg7:Duration>1400</MPEG7:Duration>
       <mpeg7:Script>http://dstc.edu.au/transcript.doc</MPEG7:Script>
       <mpeg7:Locale>Gore Hill</MPEG7:Locale>
   	<dc:Relation rdf:parseType="Resource">
     		<dcq:RelationType
				rdf:resource="http://purl.org/metadata/dublin_core_qualifiers#IsPartOf"/>
     			<rdf:value resource="http://www.dlib.org/dlib/may98/05contents.html"/>
   	</dc:Relation>
  </rdf:Description> 
</rdf:RDF>

Appendix 2: Encoding Metadata in HTML

There are two approaches to encoding metadata in HTML. One is to use a META tag for each property in the HEAD of the document. The other is to use the abbreviated RDF syntax which hides proptery values that would normally be exposed string content within tags.

META Tag Approach

Excerpt from: http://www.ietf.org/rfc/rfc2731.txt

<html>
    <head>
       <title>A Dirge</title>
       <link rel     = "schema.DC"
             href    = "http://purl.org/DC/elements/1.0/">
       <meta name    = "DC.Title"
             content = "A Dirge">
       <meta name    = "DC.Creator"
             content = "Shelley, Percy Bysshe">
       <meta name    = "DC.Type"
             content = "poem">
       <meta name    = "DC.Date"
             content = "1820">
       <meta name    = "DC.Format"
             content = "text/html">
       <meta name    = "DC.Language"
             content = "en">
     </head>
     <body>
               <p>Rough wind, that moanest loud
                 Grief too sad for song;
               Wild wind, when sullen cloud
                 Knells all the night long;
               Sad storm, whose tears are vain,
               Bare woods, whose branches strain,
               Deep caves and dreary main, -
                 Wail, for the world's wrong!</p>
     </body>
<html>

Abbreviated RDF

<html>
<head>
	<rdf:RDF
   	xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#"
  		xmlns:dc="http://purl.org/metadata/dublin_core#">
  	<rdf:Description about="http://www.dlib.org"
    	dc:Title="D-Lib Program - Research in Digital Libraries"
    	dc:Description="The D-Lib program supports the community of people
    	with research interests in digital libraries and electronic
    	publishing."
    	dc:Publisher="Corporation For National Research Initiatives"
   	dc:Date="1995-01-07"/>
	</rdf:RDF>
</head>
<body>
<P>This is a fine document.</P>
</body>
</html>

Or in the case where an RDF container such as Seq or Bag needs to be used for multiple values:

<html>
<head>
  <rdf:RDF
    xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#"
    xmlns:dc="http://purl.org/metadata/dublin_core#"> 
    <rdf:Description about="http://www.dlib.org"> 
      <dc:Creator>
			<rdf:Seq ID="CreatorsAlphabeticalBySurname"
	  			rdf:_1="Mary Andrew"
	  			rdf:_2="Jacky Crystal"/>
      </dc:Creator>
    </rdf:Description> 
  </rdf:RDF>
</head>
<body>
<P>This is a fine document.</P>
</body>
</html>

Appendix 3: Relevant XML Namespaces & Standards

Dublin Core

Dublin Core defines the 15 fields below. For more information, see: http://dublincore.org/documents/dces/

Science Namespaces

The folllowing namespaces seem to have some potential for our metadata needs, but more research is required. Ideally we will find one science and one multimedia namespace to meet all our needs that are clear standards in their respective domains that we can expect will have widespread support.

• FGDC Content Standard for Digital Geospatial Metadata (CSDGM)
• IMS Learning Resource
• Earth Science Markup Language (ESML)

Science Vocabularies

• American Geophysical Union (AGU)
• NASA DIF for GCMD
• CIESIN Indexing Vocabulary
• DLESE

Multimedia Namespaces

• MPEG-7
• Adobe XMP

Hunter-Zhan System for PNG Embedded Metadata

This system is relevant to our situation since PNG is an image format. The authors take the approach of refining the base elements in Dublin Core to describe photographs in PNG format. They utilize the Dublin Core namespace and do not create one of their own.

For more information, see: http://archive.dstc.edu.au/RDU/staff/jane-hunter/PNG/paper.html

Adobe XMP

Adobe's eXtensible Metadata Platform (XMP) is based on RDF and provides a common XML framework that standardizes the creation, processing, and interchange of document metadata across publishing workflows. Furthermore, it embeds metadata within application files, so the metadata always travels with the file. The XMP framwork has already been built into Adobe Acrobat and eventually all Adobe applications will support it. Given Adobe's market share in image and video production software, it is likely that the namespace(s) within XMP will be useful for our image and video cataloging needs.

For more information, see: http://www.adobe.com/products/xmp/main.html

DLESE

The Digital Library for Earth System Education (DLESE) has a Resource Cataloger with metadata guidelines. Of particular interest are their encoding and vocabulary recommendations on expressing coverage and subject, excerpted below.

For more information, see: http://catalog.dlese.org/catalog/cataloger/editor/best_practices.jsp

COVERAGE Definition: An area, in the form of a bounding box or point, on the surface of the Earth that a resource is about. The area is described with latitude and longitude coordinates and optionally with vertical dimensions and/or place or event names associated with the coordinate locations.

Bounding Box: A bounding box is a regular polygon that is parallel to the equator that encloses the areal extent of the location. It is used to represent, in a general way, the location of a geographic area. A bounding box is represented by two latitude and two longitude values for the edges of the box. A point is represented by repeating the latitude and longitude values.

• Latitudes run parallel to the equator and range from -90¯ to 90¯
• Longitudes run from pole to pole and range in value from -180¯ to 180¯; the prime meridian (0¯) runs through Greenwich, England.
• Enter longitude and latitude values in decimal degrees.
• Enter values in the following format ###.##, that is, to two decimal places.
• For those values less than 1 degree, use a leading zero, e.g. 0.93.
• Use positive values for longitudes east of the prime meridian (Eastern hemisphere).
• Use negative values for longitudes west of the prime meridian (Western hemisphere).
• Use positive values for latitudes north of the equator. (Northern hemisphere).
• Use negative values for latitudes south of the equator (Southern hemisphere).
• To represent a point, repeat the latitude for both the northern and southern latitudes and repeat the longitude values for both western and eastern longitudes.
• A bounding box, should cover all the sub-geographic areas within the resource. If you are cataloging a resource that encompasses the climates of Southern California, the Mediterranean, and Northern Australia, then your bounding box is approximately -30.00 (south latitude) to 35.00 (north latitude) and from -180.00 (west longitude) to 180.00 (east longitude).

Begin and End Time: Represents the beginning and ending time that pertain to the entire resource.

• Dates are indicated using a YYYY-MM-DD format.
• Time must be entered as Universal Coordinated Time in the following format: hh:mm:ss where seconds must be completed.
• Do not add UTC or Z. This will be done by the cataloging tool.
• If you enter a begin date you must enter an end date and vice versa.
• If you enter a begin time you must enter an end time and vice versa.
• If you enter a begin date/time description you must enter an end date/time description and vice versa.
• The data/time description box can be used to indicate less specific time periods such as 5000 BC or 150 million years ago, the Jurassic Period.
• If you enter a day, you must enter a year and month.
• If you enter a month, you must enter a year.
• If you enter seconds, you must enter hour and minutes.
• If you enter minutes, you must enter hour.
• To convert from local time to Universal Coordinated Time, determine the local time in military format (that is 1pm is 1300), the time of year and the location. Then use the table below and add the appropriate number of hours as indicated by a "+" or "-" sign in the table. For example, 1pm in Denver on June 30th is 19 UTC (1300 + 6 hours = 1900).

Altitude: the height above or below sea-level in meters.

• Values must be in meters.
• Do not enter commas. For example, 20,000 meters should be written as 20000.
• Use leading zeros for absolute values of less than 1. That is write 0.63 not .63.
• Positive values represent values above sea-level.
• Negative values represent values below sea-level.
• If the resource uses above or below ground measurements rather than sea-level measurements, please contact the DLESE metadata specialists for help on completing altitude.

SUBJECT Vocabulary Explanations:

Atmospheric science: The study of Earth's atmosphere including both its physical and chemical properties. Includes meteorology (weather) and some aspects of the El-Niño and La Nina phenomena, such as; precipitation patterns, atmospheric pressure, wind patterns and temperature distributions.

Chemistry: The study of a substance's composition, structure and properties, and the changes it undergoes in all three including associated energy changes. Includes molecular structure, solution and concentration, acid/base/salts, and the periodic table of elements.

Climatology: The study of the mean physical state of the atmosphere, it's statistical variation over time and space as evidenced by weather patterns over long periods of time. Includes the analysis of the causes of different climates, the presentation of climatic data and the application of climatic data to problems.

Cryology: The study of ice and snow, including glaciers, sea ice and ice shelves.

Ecology: The study of the relationships between organisms and their environments (both living and non-living) at the population, community and ecosystem levels.

Environmental science: An interdisciplinary, systems approach to understanding the link between human activities and the natural environment as well as the related impacts of humanity on individual species, ecological systems, and human health. Topics include resource use and environmental sustainability, responses of systems to anthropogenic stress; population growth; biodiversity and conservation; pollution and water quality.

Natural hazards: Events or phenomena that have the potential to inflict property damage and loss of life. This includes floods, hurricanes, earthquakes, tornadoes, volcanic eruptions and tidal waves.

Biological oceanography: The study of the flora and fauna of oceans in relation to the marine environment.

Chemical oceanography: The hydrographic and physical characterization of the ocean, including the chemical composition of seawater, salinity, the vertical and regional variation of elements, the mineralogical composition and distribution of marine sediments, and the oceanic cycle of organic and inorganic carbon and associated elements.

Physical oceanography: The study of the physical properties of Earth's oceans including temperature, sea level height, tides, currents, the movements of the sea, and the variability of these factors in relationship to the atmosphere and the ocean bottom. Also includes El Nino-related changes in sea level height, temperature and surface currents.

Appendix 4: References

Dublin Core Metadata Initiative (DCMI)

DCMI Type Vocabulary

Dublin Core Metadata Element Set

Dublin Core Qualifiers

Using Dublin Core

DCMI Box Encoding Scheme

DCMI Period Encoding Scheme

Guidance on expressing the Dublin Core within RDF

Expressing Qualified Dublin Core in RDF

Resource Description Framework (RDF)

W3C RDF Primer

Resource Description Framework (RDF) Model and Syntax Specification

Resource Description Framework (RDF) Schema Specification 1.0

Guidance on expressing the Dublin Core within the Resource Description Framework (RDF)

Science

Earth Science Markup Language

DLESE Metadata Working Group Homepage

DLESE Best Practices for Cataloging

Metadata Levels Required, Recommended and Robust Metadata

IMS Specifications Meta-data

IMS Resource Description Framework(RDF) Bindings

IMS Meta-data FAQs

NSF Funded NSDL Collection Level Metadata

Video and Images

The Application of Metadata Standards to Video Indexing

A Comparison of Schemas for Video Metadata Representation

Overview of the MPEG-7 Standard

Multimedia Metadata Standards

Adobe Systems Inc. - XMP

An Indexing, Browsing, Search and Retrieval System for Images Based on Embedded Metadata

Image Description on the Internet - CNI & OCLC

Describing and retrieving photos using RDF and HTTP

Back to Top