際際滷

際際滷Share a Scribd company logo

Enterprise Multimedia Integration and Search

O
Ozelin

playence presentation called "Enterprise Multimedia Integration and Search" for the Future Enterprise Systems Workshop, held jointly with the Future Internet Symposium in Berlin, Germany, 20th September 2010

1 of 18
Enterprise Mul-media Integra-on and Search Future Enterprise Systems September 20th, 2010 Ozelin L坦pez, Katharina Siorpaes www.playence.com
MoAvaAon ≒ With increasing bandwidth, cheaper storage of data, and improved hardware, mulAmedia content is gaining importance. ≒ 40% of worldwide tra鍖c in Internet in 2010 will be consumed by viewing and/or downloading videos ≒ In 2014, the sum of all video forms will consume 90% of tra鍖c Source: Cisco www.playence.com 20 sepAembre 2010 2
MoAvaAon ≒ As rich medium, video can transport and conserve more informaAon than text ever could. ≒ This type of content also creates new issues with respect to search, integraAon, management and preservaAon ≒ A new challenge arises, when trying to integrate text with mulAmedia www.playence.com 20 sepAembre 2010 3
MoAvaAon ≒ So far, only small parts of data comes out of automaAc analysis on mulAmedia assets ≒ The work is 鍖nally done, mostly, by humans. ≒ MulAmedia informaAon is isolated www.playence.com 20 sepAembre 2010 4
MulAmedia HolisAc View in the Enterprise ≒ Knowledge AcquisiAon process is criAcal to any semanAcally enhanced system ≒ Current informaAon systems can only rely on weak annotaAon processes for mulAmedia assets ≒ When needed, manual annotaAon tagging is performed with the help of shared vocabularies and thesauri. ≒ The level of integraAon with exisAng InformaAon Systems is limited www.playence.com 20 sepAembre 2010 5
MulAmedia HolisAc View in the Enterprise ≒ Interlinking ≒ AnnotaAon ≒ IntegraAon ≒ Search ≒ Management www.playence.com 20 sepAembre 2010 6
AnnotaAon Process ≒ The annotaAon process can be done automaAcally by the system or manually. The accuracy and precision of this process depends on the type of content. ≒ Textual annotaAon is done automaAcally, using Ontologies and NLP techniques to pinpoint textual references to concepts and instances in the source. ≒ Several domain ontologies can be used to this purpose, providing a mulA-足view perspecAve on the same resource. www.playence.com 20 sepAembre 2010 7
AnnotaAon Process ≒ In the case of video or audio, automaAc analysis has less accuracy. ≒ ASR can done the work up to some extent (60-足80%) ≒ For video or image analysis, high-足level features can be obtained, like face detecAon, detecAon of objects, daylight classi鍖caAon and other basic features. ≒ In these cases, collaboraAon human annotaAon must be supported. www.playence.com 20 sepAembre 2010 8
AnnotaAon Process ≒ In the case of video or audio, automaAc analysis has less accuracy. ≒ ASR can done the work up to some extent (60-足80%) ≒ For video or image analysis, high-足level features can be obtained, like face detecAon, detecAon of objects, daylight classi鍖caAon and other basic features. ≒ In these cases, collaboraAon human annotaAon must be supported. www.playence.com 20 sepAembre 2010 9
IntegraAon ≒ AnnotaAon process will leave a set of resources linked to the same semanAc content ≒ Data can easily be located, mashed-足up and displayed regardless its original source. ≒ IntegraAon in playence Media empowers the user to locate a meeAng recording and being able to 鍖nd related videos, audios, pictures and documents. ≒ Once annotated, everything can be queried using the same model. www.playence.com 20 sepAembre 2010 10
IntegraAon www.playence.com 20 sepAembre 2010 11
Search ≒ playence Media performs semanAc search, using annotaAons and applying faceted search and semanAc navigaAon to narrow the set of results ≒ When searching, playence Media makes use of Natural Language Processing techniques like lemmaAzaAon or spell check. ≒ SemanAc features are used in query expansion, like synonym expansion through SKOS, or generalizaAon-足 specializaAon expansion, using the is-足a relaAonship and instances from concepts involved, or using more complex relaAons in query expansion. www.playence.com 20 sepAembre 2010 12
Search www.playence.com 20 sepAembre 2010 13
playence Media Intelligent mulAmedia management including: automaAc semanAc annotaAon, interlinking, in-足video search and browsing. ≒ Automa-c mul--足language knowledge extrac-on LocaAon of relevant key-足words and domain knowledge through batch ASR. E鍖cient video annotaAon for later use. ≒ In-足video Search Land a search at the exact second where the relevant content is being played. Land a search at the exact second where the relevant actor is talking. ≒ Mul-linguality Support for a wide variety of languages. Complex cross-足language query and mulAmedia asset retrieval. ≒ Mul-format support AVI, MPEG, FLV, etc. www.playence.com 20 sepAembre 2010 14
playence Media ≒ Dynamic ra-ng On-足the-足go signalling on the most interesAng porAons of a rich media asset. Beher locaAon of releant content and improve search results. ≒ Manual annota-on re鍖nement Re鍖nement of automaAc annotaAons. AddiAon of annotaAons. ≒ Browsing and asset naviga-on Powerful browsing engine to keep sight of huge amounts of media content. Eases 鍖nding, discovering and accessing the required media resources. ≒ Rela-onship viewer Snapshot view of a porAon of the subjacent data model. Document set 鍖ltering by way of query expansion and reducAon. www.playence.com 20 sepAembre 2010 15
Further Steps ≒ The challenges associated with mulAmedia integraAon in the enterprise are manifold ≒ Relevance and in-足video search ≒ Ontology evoluAon from customer perspecAve ≒ Work鍖ow and collaboraAve processes are needed. ≒ Interlinking Linked Open Data Linked Closed Data www.playence.com 20 sepAembre 2010 16
Conclusion ≒ Media is going to be the next enterprise communicaAon mechanism ≒ Media is a bitch! We need completely new ways of manage it Business is beyond text, but technology is not ≒ Providing a holisAc view empowers enterprises to manage mulAmedia assets ≒ This holisAc view comprises heavily informaAon related processes: annotaAon, integraAon, search, interlinking ≒ playence Media comes to the playground to help companies dealing with these challenges. www.playence.com 20 sepAembre 2010 17
QuesAons ? We understand and integrate your media content. www.playence.com

More Related Content

Enterprise Multimedia Integration and Search

  • 1. Enterprise Mul-media Integra-on and Search Future Enterprise Systems September 20th, 2010 Ozelin L坦pez, Katharina Siorpaes www.playence.com
  • 2. MoAvaAon ≒ With increasing bandwidth, cheaper storage of data, and improved hardware, mulAmedia content is gaining importance. ≒ 40% of worldwide tra鍖c in Internet in 2010 will be consumed by viewing and/or downloading videos ≒ In 2014, the sum of all video forms will consume 90% of tra鍖c Source: Cisco www.playence.com 20 sepAembre 2010 2
  • 3. MoAvaAon ≒ As rich medium, video can transport and conserve more informaAon than text ever could. ≒ This type of content also creates new issues with respect to search, integraAon, management and preservaAon ≒ A new challenge arises, when trying to integrate text with mulAmedia www.playence.com 20 sepAembre 2010 3
  • 4. MoAvaAon ≒ So far, only small parts of data comes out of automaAc analysis on mulAmedia assets ≒ The work is 鍖nally done, mostly, by humans. ≒ MulAmedia informaAon is isolated www.playence.com 20 sepAembre 2010 4
  • 5. MulAmedia HolisAc View in the Enterprise ≒ Knowledge AcquisiAon process is criAcal to any semanAcally enhanced system ≒ Current informaAon systems can only rely on weak annotaAon processes for mulAmedia assets ≒ When needed, manual annotaAon tagging is performed with the help of shared vocabularies and thesauri. ≒ The level of integraAon with exisAng InformaAon Systems is limited www.playence.com 20 sepAembre 2010 5
  • 6. MulAmedia HolisAc View in the Enterprise ≒ Interlinking ≒ AnnotaAon ≒ IntegraAon ≒ Search ≒ Management www.playence.com 20 sepAembre 2010 6
  • 7. AnnotaAon Process ≒ The annotaAon process can be done automaAcally by the system or manually. The accuracy and precision of this process depends on the type of content. ≒ Textual annotaAon is done automaAcally, using Ontologies and NLP techniques to pinpoint textual references to concepts and instances in the source. ≒ Several domain ontologies can be used to this purpose, providing a mulA-足view perspecAve on the same resource. www.playence.com 20 sepAembre 2010 7
  • 8. AnnotaAon Process ≒ In the case of video or audio, automaAc analysis has less accuracy. ≒ ASR can done the work up to some extent (60-足80%) ≒ For video or image analysis, high-足level features can be obtained, like face detecAon, detecAon of objects, daylight classi鍖caAon and other basic features. ≒ In these cases, collaboraAon human annotaAon must be supported. www.playence.com 20 sepAembre 2010 8
  • 9. AnnotaAon Process ≒ In the case of video or audio, automaAc analysis has less accuracy. ≒ ASR can done the work up to some extent (60-足80%) ≒ For video or image analysis, high-足level features can be obtained, like face detecAon, detecAon of objects, daylight classi鍖caAon and other basic features. ≒ In these cases, collaboraAon human annotaAon must be supported. www.playence.com 20 sepAembre 2010 9
  • 10. IntegraAon ≒ AnnotaAon process will leave a set of resources linked to the same semanAc content ≒ Data can easily be located, mashed-足up and displayed regardless its original source. ≒ IntegraAon in playence Media empowers the user to locate a meeAng recording and being able to 鍖nd related videos, audios, pictures and documents. ≒ Once annotated, everything can be queried using the same model. www.playence.com 20 sepAembre 2010 10
  • 11. IntegraAon www.playence.com 20 sepAembre 2010 11
  • 12. Search ≒ playence Media performs semanAc search, using annotaAons and applying faceted search and semanAc navigaAon to narrow the set of results ≒ When searching, playence Media makes use of Natural Language Processing techniques like lemmaAzaAon or spell check. ≒ SemanAc features are used in query expansion, like synonym expansion through SKOS, or generalizaAon-足 specializaAon expansion, using the is-足a relaAonship and instances from concepts involved, or using more complex relaAons in query expansion. www.playence.com 20 sepAembre 2010 12
  • 13. Search www.playence.com 20 sepAembre 2010 13
  • 14. playence Media Intelligent mulAmedia management including: automaAc semanAc annotaAon, interlinking, in-足video search and browsing. ≒ Automa-c mul--足language knowledge extrac-on LocaAon of relevant key-足words and domain knowledge through batch ASR. E鍖cient video annotaAon for later use. ≒ In-足video Search Land a search at the exact second where the relevant content is being played. Land a search at the exact second where the relevant actor is talking. ≒ Mul-linguality Support for a wide variety of languages. Complex cross-足language query and mulAmedia asset retrieval. ≒ Mul-format support AVI, MPEG, FLV, etc. www.playence.com 20 sepAembre 2010 14
  • 15. playence Media ≒ Dynamic ra-ng On-足the-足go signalling on the most interesAng porAons of a rich media asset. Beher locaAon of releant content and improve search results. ≒ Manual annota-on re鍖nement Re鍖nement of automaAc annotaAons. AddiAon of annotaAons. ≒ Browsing and asset naviga-on Powerful browsing engine to keep sight of huge amounts of media content. Eases 鍖nding, discovering and accessing the required media resources. ≒ Rela-onship viewer Snapshot view of a porAon of the subjacent data model. Document set 鍖ltering by way of query expansion and reducAon. www.playence.com 20 sepAembre 2010 15
  • 16. Further Steps ≒ The challenges associated with mulAmedia integraAon in the enterprise are manifold ≒ Relevance and in-足video search ≒ Ontology evoluAon from customer perspecAve ≒ Work鍖ow and collaboraAve processes are needed. ≒ Interlinking Linked Open Data Linked Closed Data www.playence.com 20 sepAembre 2010 16
  • 17. Conclusion ≒ Media is going to be the next enterprise communicaAon mechanism ≒ Media is a bitch! We need completely new ways of manage it Business is beyond text, but technology is not ≒ Providing a holisAc view empowers enterprises to manage mulAmedia assets ≒ This holisAc view comprises heavily informaAon related processes: annotaAon, integraAon, search, interlinking ≒ playence Media comes to the playground to help companies dealing with these challenges. www.playence.com 20 sepAembre 2010 17
  • 18. QuesAons ? We understand and integrate your media content. www.playence.com