; Media data
Learning Center
Plans & pricing Sign in
Sign Out
Your Federal Quarterly Tax Payments are due April 15th Get Help Now >>

Media data


  • pg 1
									Media 2.0 – 媒体新技术革命?

       李世鹏 博士

      微软亚 洲研究院
    高级 研究员 、院长 助理
     网络 多媒体组 主任
• Introduction
   – What’s happening?
   – Web 2.0
• Media 2.0 – my media 5-D’s
   –   Democratized life cycle
   –   Data-driven value chain
   –   Decoupled system
   –   Decomposed contents
   –   De-centralized business model
• Demos
       Introduction – What’s happening?
•   youtube.com
     – User contributed contents (UCC)
     – User community
     – Blogging/tagging/commenting/rating
•   blinx.com
     – Content aggregation and video search
•   PPLive/Coolstream
     – P2P streaming
•   Adobe’s Apollo project
     – Flash player jumps beyond browser
•   Apple iPod
     – iTune services & podcasting
•   Microsoft Zune
     – Connected media experience and community sharing
•   Google Adsense
     – Long-tail advertisement
               Introduction – Web 2.0
•   Web as a platform
•   Harnessing collective intelligence
•   Data is the next Intel Inside
•   End of the software release cycle
•   Lightweight programming model
•   Software above the level of a single device
         Tom O’Reilly’s “What is Web 2.0”
        Media 2.0 – My Media 5-D’s
•   Democratized media life cycle
•   Data-driven media value chain
•   Decoupled media system
•   Decomposed media contents
•   Decentralized media business model
Media Life Cycle


          Democratized Media Life Cycle
•   Grassroots contributions
     –   Capturing/Authoring
     –   Publishing/Distribution/Sharing
     –   Referring/Linking
     –   Commenting/Rating
•   Applications
     –   User contributed media contents
     –   Video Wiki
     –   P2P download/ streaming
     –   Media tagging
     –   Community recommendation/sharing
•   Technologies: tools for users to easy capture and author media contents
     – LazyMedia
     – AutoMovie
         Data-driven Media Value Chain
•   Rich data beyond the media data itself
     – Media data: media content itself
     – User data: user comments/rating/tagging
     – Metadata: information about the media contents, such as location, time,
       origin, content descriptors, etc., either through human interaction or
       automatic generation
     – Linkage data: links between media contents to turn individual media
       islands into media networks
     – Hits data: Media popularity

•   Media value beyond the media contents with rich data/context
     – Facilitate media management/search
     – Create media social networks
     – Mine user profile/preferences
                Decoupled Media System
•   On-demand reconfigurable and adaptive media processing pipe

•   LiquidSilver project at MSRA
     – Mimic a Unix-command like architecture
     – Break a whole media process into basic units
     – Re-use units as much as possible
•   Features
     – Lightweight modules
          • All processing units are independent
          • Connected by data flow
     – Pluggable components
          • Easy replacement of existing units
          • Online downloadable
          • Short release cycle
     – Reconfigurable through web services
     – Automatically adapt to networks and user device capability
            Decomposed Media Contents
•   Decompose media contents into small unit for easy analysis and access
•   Media analogy to text
     – To enable media analysis as easy as text,
     – scenes as sections, shots as paragraphs, frames as words
•   Content-based search
     – Traditional video search
          • based on search “direct text” – surrounding text context
          • “Direct text” != Content
     – Content-based video indexing
          • From “direct text” to “hidden text”
          • Convert low-level features to semantic terms/keywords/ontology
•   Content-based video presentation
     – Enable fast video content browsing on the Web
•   Inner media hyperlink
     – Link into media contents rather than to a whole clip
    Decentralized Media Business Model
•   Centralized big-head model è Distributed long-tail model
     – Google adsense, Microsoft AdLive
•   Value generated by media content only è Value-added by related rich media
     – Use bahavior profiling, targeted ads
•   Central server model è superdistribution model
     – Take advantages of social networks and micro-payment
     – “Super-girl” phenomenon
•   The direct value of media content may be diluted even offered for free
     – The greater value may be recollected through other means above
•   Technical challenges:
     – Targeted Ads in media space
     – Non-intrusive Ads insertion
LiquidSilver Demo
Video Search & Presentation Demo
•   Media 2.0 could potentially bring a revolution in internet media
•   The 5-D features of Media 2.0 can be further summarized into one

                 From Grassroots to Grassroots

•   Beyond technical challenges
     – Politics and local laws
     – Government engagement

To top