Semantic multimedia panel at WWW2007
More rough notes from the WWW2007: this time i am listening to the panel on The Role of Multimedia Metadata Standards in a (Semantic) Web 3.0 moderated and organized by Raphael Troncy and Susanne Bol.
Raphael Troncy: Introduction
- jungle of metadata formats for MM: MPEG7, IPTC, XMP
- but also tagging: Youtube, Flickr, Lastfm
- and semantic web: OWL, RDF, FOAF+SKOS
- yet, there are interconnections
Lynda Hardman: Is the Message in the Media?
- we want tags with meaning
- video segments as building blocks in the process of
- images of penal speakers are very heterogeneous
- finding images should bring images in a common representation
- video segments: speakers are in different contexts
- continuity and consistency (light, sound, movement)
Sean Bechhofer (U of Manchester): Semantic Web for MM
- SW is about annotation, integration, and inference
- using existing infrastructures to have migration paths
- MPEG7 lacks inference, but does ann’ and int’
John Smith (IBM): Real problem of Bridging the MM “Semantic Gap”
- multimedia/video is big in terms of traffic and storage
- what is the problem? focus on search
- video requires metadata (as opposed to text)
- beyond content chaos, now: metadata chaos
- different approaches along the long tail:
- manual cataloging by pros for popular items
- automated tagging for deep/raw footage
- social tagging by users: personal content
- “semantic gap” has been seen as a machine learning problem
- better: leverage digital masses/crowd
Mor Naaman (Yahoo): Developers are people, too
- people are strange, don’t care about semantics
- but people tag: simple (Rashmi Sinha), powerful, motivated
- generate semantics from tags, e.g. Flickr clusters
- metadata helps, too: geocoding
- “Extracting Tag Semantics”, SIGIR2007
- req for dev app: simple, powerful, tied to motivations
- so far: (GEO-)RSS, KML, GPX, EXIF (read only)
- and Flickr machine tags: subject-predicate-object
- … simple and powerful
- problem: namespace and semantics not defined
- this is up to researchers
- welcome to the emerging semantics web
Discussion
- media object including metadata (W3C Web API WG)
- a designated microformat could do this
- SW is more than a triple in a free form field
- tags of YouTube and Flickr describe resources better
than any format / standard would have done
- most work of feature analysis focuses on sunsets
- is Flickr’s machine tag a new microformat for RDF?
- RDF might not be fashionable, but it’s “bloody simple”
- RDF enables mashing-up (see open/linked data)
- Flickr’s machine tags can be mapped to RDF (flickcurl)
- get rid of feature analyis?
- query by example seems not to be centrally popular
- “Future of the Web” by Nat Torkington
About this entry
You’re currently reading “Semantic multimedia panel at WWW2007,” an entry on Anarchitect
- Published:
- 5.10.07 / 11pm
- Category:
- education, english, technology
- Tags:
No comments
Jump to comment form | comments rss [?] | trackback uri [?]