Introduction to metadata

1
Introduction

• Introduction to metadata
• Metadata you already use - ID3 tags
• Metadata you might want to use - MPEG7
• Flexible search capabilities
• Storing complex information in multimedia
  databases - MMDBs

2
What is metadata?

• Metadata literally means beyond data or
  second-order data
• It is information about information in the
  digital world it is usually extra information
  that supports the main channel of data
• The title of this slide is both data and
  metadata, depending on your point of view

3
Examples of metadata

• One kind of metadata you will know well is the
  ID3 tag in MP3 files
• Other music metadata could include
• Lyrics
• Artist web links

4
Examples of metadata

• Movie subtitles are a good example of metadata
• However, a well-annotated movie will also have
  other information in a digitally searchable
  format
• Cast and crew members (immutable)
• Synopsis of the action (e.g. the tower fades
  into the mist), particularly useful for special
  needs users (mutable)
• Scene content description keywords (e.g. tower,
  garden, moonlight, heroine, mist) for
  search engines
• As we can see, high quality metadata is expensive

5
Standards for metadata MPEG-7

• MPEG-7 is the Multimedia Content Description
  Interface
• It represents information about content, not the
  content itself
• It uses XML to store metadata
• It is independent of other MPEG encoding
  standards but is expected to be used with them
• It can be linked to timecodes to provide lyrics
  or subtitles during the progress of associated
  media

6
MPEG-7 example
Reference http//www.semanticmetadata.net/images/
screenshots/caliph-03.png SemanticMetadata.net
19/02/07

• Shape information may be added by hand
• Shape information is stored within the MPEG-7
  document
• This enables query-by-shape and shape comparison

7
Problems with metadata

• Some companies may not produce metadata because
  of high costs
• Private users may not produce complex metadata
  because it is very time-consuming
• MPEG-7 is complicated for non-professional users
• Automatic tools for creating high-quality
  metadata are not yet available

8
Problems with metadata

• Some metadata depends on context thus different
  people would create very different metadata
  streams for the same media
• Context-sensitive metadata may be misinterpreted
• Manually created metadata may be more or less
  valuable in the near future no-one knows
• Techniques like query-by-example allow users to
  find pictures, videos and music without the need
  for metadata (e.g. query-by-humming)

9
Multimedia Databases

• Any modern database can store more than just text
  and numbers
• Some specialise in storing media
• We may be able to search a MMDB by normal query
• Get all records where creation date is July
  2008
• Or we may use a query-by-example
• Sketch a shape to find shapes that are similar
• Provide a photo to find images containing a
  matching face
• Hum a tune to find music with similar passages

10
Summary

• We have seen the basic ideas of metadata
• ID3 tags provide a well known example of metadata
• MPEG7 is a complex but powerful method of
  storing, organising and linking metadata
• Digital media encourages highly varied search
  capabilities
• We can store complex information in multimedia
  databases, so MMDBs are good for storing metadata
  and for QbE

11
Links and reading

• MPEG-7 Multimedia Content Description
  Standardhttp//www.chiariglione.org/mpeg/tutori
  als/papers/IEEEMM_mp7overview_withcopyrigth.pdf
• Query by Humming http//en.wikipedia.org/wiki/Que
  ry_by_humming
• ID3v2 Homepage http//www.id3.org/Home
• Windows Media Metafileshttp//msdn2.microsoft.co
  m/en-gb/library/aa385262.aspx
• Eptascape demos http//www.eptascape.com/products
  /demoflv.htm
• SemanticMetadata.net screenshotshttp//www.seman
  ticmetadata.net/screenshots/
• Cory Doctorow comments http//www.well.com/docto
  row/metacrap.htm