SlideShare a Scribd company logo
1 of 28
Download to read offline
Speech & Language
                                                         Technology
                                                  A.J.vanHessen@ewi.utwente.nl


                               	
  

Searching	
  in	
  spoken	
  words
 Disclosure of recorded content in MediaMosa
            
                              	

            SURFnet Relatiedagen 2010
            Noordwijkerhout December 9, 2010
•  Introduction	
  
   –  Why	
  speech	
  is	
  so	
  important	
  
   –  What	
  is	
  HLT?	
  


•  Working	
  applications:	
  
   –  Self-­‐service	
  (Internet	
  	
  Telephony)	
  
   –  Searching	
  in	
  recorded	
  audiovisual	
  recordings	
  

•  Demonstrations	
  
Humans as speaking creatures
The	
  start	
  of	
  the	
  human	
  
speech	
  started	
  some	
  
100.000	
  years	
  ago.	
  
Before,	
  the	
  shape	
  of	
  the	
  
vocal	
  track	
  was	
  not	
  “ready”	
  
for	
  the	
  modern	
  speech.	
  The	
  
larynx	
  was	
  situated	
  too	
  
high,	
  something	
  you	
  can	
  see	
  
with	
  chimps.	
  
Humans as writing
                                  creatures

Sumer	
  (3300	
  AD,	
  Mesopotamia)	
  is	
  
probably	
  the	
  oldest	
  written	
  language.	
  




                                                                                 NU	
  




           -­‐100.000	
                                 -­‐10.000	
     -­‐3300	
  
              speech	
                                  farming	
          text	
  
What	
  is	
  HLT?	
  

•  Human	
  Language	
  Technology	
  is	
  the	
  
   technology	
  that	
  mimics	
  the	
  human	
  language	
  
   capacity.	
  




    speech                text                 sign
Redundancy	
  
•  Vlgones	
  een	
  oznrdeeok	
  op	
  een	
  Eglnese	
  
   uvinretsiet	
  mkaat	
  het	
  neit	
  uitin	
  wlkee	
  
   vloogdre	
  de	
  ltteers	
  in	
  een	
  wrood	
  saatn,	
  het	
  
   einge	
  watblegnaijrk	
  is	
  is	
  dat	
  de	
  eretse	
  en	
  
   de	
  ltaatse	
  ltteer	
  op	
  de	
  jiutsepatals	
  saatn.	
  
   De	
  rset	
  van	
  de	
  ltteers	
  mgoen	
  wllikueirg	
  
   gpletaastwdoren	
  en	
  je	
  knut	
  vrelvogens	
  
   gwoeon	
  lzeen	
  wat	
  er	
  saatt.	
  Dit	
  kmotodmat	
  
   we	
  neit	
  ekle	
  ltteer	
  op	
  zcih	
  lzeen	
  maar	
  het	
  
   wrood	
  als	
  gheeel.	
  
Pensez a
         ceque v
                 s avez f
                         nit   et dema
                                      n   dez vs i
                                                    est coq
                                                             ue   3 ai
Dialogue	
  systems	
  (telephony,	
  real	
  time,	
  limited	
  complexity)	
  
Disclosure	
  systems	
  (high	
  quality	
  audio,	
  ofQline,	
  complex)	
  

WORKING	
  APPLICATIONS	
  
Spoken	
  Document	
  
             ContactCenter	
  
                                                 Retrieval                    Information	
  
                                                                                Retrieval

 Voice	
  


                                 HLT
                                                                            Natural	
  Language	
  
                                                                                 Search	
  


Web	
  




                                                                              Mens-­‐Machine-­‐
                                                      Emotie	
  detectie:
Mobile	
                         Dictation             Lachen/Huilen          Communicatie
Companies	
  using	
  speech	
  technology	
  
How	
  may	
  I	
  help	
  you	
  
                                                    Why are they
   Who is calling?                                    calling?
                                                 Classification based on
Identification via ZIP-
                                                 the recognition of the
code and house number
                                                 question: “how may I
                                                 help you”
OrganisaCons	
  using	
  	
  
speech	
  technology	
  
Disclosure	
  of	
  audiovisual	
  archives	
  

•  The	
  number	
  of	
  AV-­‐archives	
  on	
  the	
  Internet	
  
   increases	
  rapidly	
  
•  Archiving	
  is	
  not	
  enough:	
  disclosure	
  and	
  
   reusing	
  is	
  required!	
  
•  The	
  use	
  of	
  HLT	
  is	
  needed	
  (humans	
  cost	
  too	
  
   much).	
  
Digitalized (historic)        LVSR	
     Digital recorded collections
                      collections




WFH




                               H.M. Koningin
      Second feministic wave     Wilhelmina

                                Buchenwald




                               Memories of Indonesia
Searching	
  in	
  historic	
  radio	
  recordings:	
  
                Radio	
  Oranje	
  
Oral	
  History:	
  Buchenwald	
  
Oral	
  History:	
  Brandgrens,	
  RoKerdam	
  




   10	
  getuigen	
  van	
  het	
  bombardement	
  van	
  Rotterdam	
  (mei	
  ‘40)	
  vertellen	
  hun	
  
   verhaal.	
  TST	
  wordt	
  gebruikt	
  om	
  in	
  de	
  getuigenissen	
  te	
  zoeken.	
  
Searching	
  in	
  the	
  radio	
  interviews	
  of	
  WFH	
  
Searching	
  in	
  46	
  interview	
  collecCons:	
  
    getuigenverhalen	
  (600	
  hour)	
  
Searching	
  in	
  500	
  interviews	
  in	
  CroaCa	
  
CroMe	
  -­‐	
  Audio	
  Search	
  
                                                   Searching	
  for:	
  
                                                   commandant	
  




   Phrase	
  
 boundaries	
  




   Found	
  word	
            5	
  fragments	
  
(5x	
  commandant)	
                 found	
  
CroMe	
  -­‐	
  Audio	
  Search	
  

                        Search	
  word	
         traumas	
  

                                       Language	
  

                           found	
  
PoliCcal	
  meCngs	
  
Parliament	
  

                                                     transcriptions	
  




	
  Gisteren	
  	
  was	
  	
  er	
  	
  een	
  	
  bespreking	
  	
  	
  	
  ivm	
  	
  	
  	
  de	
  	
  betrekkingen	
  tussen	
  Nederland	
  en	
  Vlaanderen	
  
RecogniCon	
  of	
  lectures	
  

•  Record	
  the	
  speech	
  
•  Record	
  the	
  PPT	
  
•  Recognise	
  the	
  speech	
  
•  Use	
  the	
  display	
  time	
  of	
  each	
  slide	
  as	
  THE	
  
   time	
  unit	
  
•  Use	
  the	
  recognised	
  speech	
  as	
  keywords	
  for	
  
   each	
  slide	
  
Searching	
  in	
  news	
  broadcasts	
  
Metadata	
  -­‐	
  Language	
  model	
  
                        Text in the slide(s)
Lectures handouts




                                               Language model




Environmental texts
QuesCons?	
  

More Related Content

What's hot

Tugasan 4 kpt 6044
Tugasan 4 kpt 6044Tugasan 4 kpt 6044
Tugasan 4 kpt 6044nurrasyiqah
 
Major presentation
Major presentationMajor presentation
Major presentationStuti Shukla
 
Contemporary Arts by Teena
Contemporary Arts by TeenaContemporary Arts by Teena
Contemporary Arts by Teenateenaymulanday
 
The Future of Search (Keynote at I-Know 2010)
The Future of Search (Keynote at I-Know 2010)The Future of Search (Keynote at I-Know 2010)
The Future of Search (Keynote at I-Know 2010)marti_hearst
 
Technology used by the Deaf community
Technology used by the Deaf communityTechnology used by the Deaf community
Technology used by the Deaf communityMartin Vu
 
Technologies for Deaf People
Technologies for Deaf PeopleTechnologies for Deaf People
Technologies for Deaf PeoplePamil Gurung
 
Lesson 2 the attkinson-shiffrin's multi-store model 2012 sh
Lesson 2   the attkinson-shiffrin's multi-store model 2012 shLesson 2   the attkinson-shiffrin's multi-store model 2012 sh
Lesson 2 the attkinson-shiffrin's multi-store model 2012 shcoburgpsych
 
Collecting and Evaluating Speech Recognition Corpora for Nine Southern Bantu ...
Collecting and Evaluating Speech Recognition Corpora for Nine Southern Bantu ...Collecting and Evaluating Speech Recognition Corpora for Nine Southern Bantu ...
Collecting and Evaluating Speech Recognition Corpora for Nine Southern Bantu ...Guy De Pauw
 
Multi store model of memory
Multi store model of memoryMulti store model of memory
Multi store model of memoryleannacatherina
 
Natural Language Processing and Machine Learning
Natural Language Processing and Machine LearningNatural Language Processing and Machine Learning
Natural Language Processing and Machine LearningKarthik Sankar
 
Tugasan 4 : Penilaian Kendiri Teks
Tugasan 4 : Penilaian Kendiri TeksTugasan 4 : Penilaian Kendiri Teks
Tugasan 4 : Penilaian Kendiri Teksidahazira
 
Language and Intelligence
Language and IntelligenceLanguage and Intelligence
Language and Intelligencebutest
 

What's hot (15)

Tugasan 4 kpt 6044
Tugasan 4 kpt 6044Tugasan 4 kpt 6044
Tugasan 4 kpt 6044
 
Major presentation
Major presentationMajor presentation
Major presentation
 
Contemporary Arts by Teena
Contemporary Arts by TeenaContemporary Arts by Teena
Contemporary Arts by Teena
 
The Future of Search (Keynote at I-Know 2010)
The Future of Search (Keynote at I-Know 2010)The Future of Search (Keynote at I-Know 2010)
The Future of Search (Keynote at I-Know 2010)
 
Iasa Presentatie
Iasa PresentatieIasa Presentatie
Iasa Presentatie
 
Technology used by the Deaf community
Technology used by the Deaf communityTechnology used by the Deaf community
Technology used by the Deaf community
 
Technologies for Deaf People
Technologies for Deaf PeopleTechnologies for Deaf People
Technologies for Deaf People
 
Lesson 2 the attkinson-shiffrin's multi-store model 2012 sh
Lesson 2   the attkinson-shiffrin's multi-store model 2012 shLesson 2   the attkinson-shiffrin's multi-store model 2012 sh
Lesson 2 the attkinson-shiffrin's multi-store model 2012 sh
 
Collecting and Evaluating Speech Recognition Corpora for Nine Southern Bantu ...
Collecting and Evaluating Speech Recognition Corpora for Nine Southern Bantu ...Collecting and Evaluating Speech Recognition Corpora for Nine Southern Bantu ...
Collecting and Evaluating Speech Recognition Corpora for Nine Southern Bantu ...
 
Multi store model of memory
Multi store model of memoryMulti store model of memory
Multi store model of memory
 
Natural Language Processing and Machine Learning
Natural Language Processing and Machine LearningNatural Language Processing and Machine Learning
Natural Language Processing and Machine Learning
 
Prototyping for Communication
Prototyping for CommunicationPrototyping for Communication
Prototyping for Communication
 
Final memory
Final memoryFinal memory
Final memory
 
Tugasan 4 : Penilaian Kendiri Teks
Tugasan 4 : Penilaian Kendiri TeksTugasan 4 : Penilaian Kendiri Teks
Tugasan 4 : Penilaian Kendiri Teks
 
Language and Intelligence
Language and IntelligenceLanguage and Intelligence
Language and Intelligence
 

Similar to MediaMosa Transcription technology

New life for old media - Investigations into Speech Synthesis and Deep Learni...
New life for old media - Investigations into Speech Synthesis and Deep Learni...New life for old media - Investigations into Speech Synthesis and Deep Learni...
New life for old media - Investigations into Speech Synthesis and Deep Learni...Sound and Vision R&D
 
Matt Feiszli at AI Frontiers : Video Understanding
Matt Feiszli at AI Frontiers : Video UnderstandingMatt Feiszli at AI Frontiers : Video Understanding
Matt Feiszli at AI Frontiers : Video UnderstandingAI Frontiers
 
Speech recognition techniques
Speech recognition techniquesSpeech recognition techniques
Speech recognition techniquessonukumar142
 
Finding Ostriches in the Courtroom
Finding Ostriches in the CourtroomFinding Ostriches in the Courtroom
Finding Ostriches in the CourtroomChristopher Collins
 
Natural Language Processing with Python
Natural Language Processing with PythonNatural Language Processing with Python
Natural Language Processing with PythonBenjamin Bengfort
 
Speech Technology Overview
Speech Technology OverviewSpeech Technology Overview
Speech Technology Overviewamr0mt
 
Laura Welcher - The Rosetta Project and The Language Commons
Laura Welcher - The Rosetta Project and The Language CommonsLaura Welcher - The Rosetta Project and The Language Commons
Laura Welcher - The Rosetta Project and The Language Commonslongnow
 
NOVA Data Science Meetup 1/19/2017 - Presentation 2
NOVA Data Science Meetup 1/19/2017 - Presentation 2NOVA Data Science Meetup 1/19/2017 - Presentation 2
NOVA Data Science Meetup 1/19/2017 - Presentation 2NOVA DATASCIENCE
 
Communication skills about language
Communication skills about languageCommunication skills about language
Communication skills about languageIhsan Ullah Khan
 
Computer mediated communication (cmc) as a subject
Computer mediated communication (cmc) as a subjectComputer mediated communication (cmc) as a subject
Computer mediated communication (cmc) as a subjectuhandai
 
How Does Text-to-Speech Work?
How Does Text-to-Speech Work? How Does Text-to-Speech Work?
How Does Text-to-Speech Work? Listnr
 
Natural_Language_Processing_1.ppt
Natural_Language_Processing_1.pptNatural_Language_Processing_1.ppt
Natural_Language_Processing_1.ppttestbest6
 
Natural language processing (NLP)
Natural language processing (NLP) Natural language processing (NLP)
Natural language processing (NLP) ASWINKP11
 
Class2 - The Role of Communication
Class2 - The Role of CommunicationClass2 - The Role of Communication
Class2 - The Role of CommunicationNathacia Lucena
 
Digital speech processing lecture1
Digital speech processing lecture1Digital speech processing lecture1
Digital speech processing lecture1Samiul Parag
 
Speech Recognition Technology
Speech Recognition TechnologySpeech Recognition Technology
Speech Recognition TechnologyAamir-sheriff
 

Similar to MediaMosa Transcription technology (20)

Amadou
AmadouAmadou
Amadou
 
New life for old media - Investigations into Speech Synthesis and Deep Learni...
New life for old media - Investigations into Speech Synthesis and Deep Learni...New life for old media - Investigations into Speech Synthesis and Deep Learni...
New life for old media - Investigations into Speech Synthesis and Deep Learni...
 
Matt Feiszli at AI Frontiers : Video Understanding
Matt Feiszli at AI Frontiers : Video UnderstandingMatt Feiszli at AI Frontiers : Video Understanding
Matt Feiszli at AI Frontiers : Video Understanding
 
Speech recognition techniques
Speech recognition techniquesSpeech recognition techniques
Speech recognition techniques
 
Finding Ostriches in the Courtroom
Finding Ostriches in the CourtroomFinding Ostriches in the Courtroom
Finding Ostriches in the Courtroom
 
Intro
IntroIntro
Intro
 
Intro
IntroIntro
Intro
 
Natural Language Processing with Python
Natural Language Processing with PythonNatural Language Processing with Python
Natural Language Processing with Python
 
Speech Technology Overview
Speech Technology OverviewSpeech Technology Overview
Speech Technology Overview
 
Laura Welcher - The Rosetta Project and The Language Commons
Laura Welcher - The Rosetta Project and The Language CommonsLaura Welcher - The Rosetta Project and The Language Commons
Laura Welcher - The Rosetta Project and The Language Commons
 
NOVA Data Science Meetup 1/19/2017 - Presentation 2
NOVA Data Science Meetup 1/19/2017 - Presentation 2NOVA Data Science Meetup 1/19/2017 - Presentation 2
NOVA Data Science Meetup 1/19/2017 - Presentation 2
 
Communication skills about language
Communication skills about languageCommunication skills about language
Communication skills about language
 
Computer mediated communication (cmc) as a subject
Computer mediated communication (cmc) as a subjectComputer mediated communication (cmc) as a subject
Computer mediated communication (cmc) as a subject
 
How Does Text-to-Speech Work?
How Does Text-to-Speech Work? How Does Text-to-Speech Work?
How Does Text-to-Speech Work?
 
Natural_Language_Processing_1.ppt
Natural_Language_Processing_1.pptNatural_Language_Processing_1.ppt
Natural_Language_Processing_1.ppt
 
Natural language processing (NLP)
Natural language processing (NLP) Natural language processing (NLP)
Natural language processing (NLP)
 
Class2 - The Role of Communication
Class2 - The Role of CommunicationClass2 - The Role of Communication
Class2 - The Role of Communication
 
Digital speech processing lecture1
Digital speech processing lecture1Digital speech processing lecture1
Digital speech processing lecture1
 
Speech Recognition Technology
Speech Recognition TechnologySpeech Recognition Technology
Speech Recognition Technology
 
intro.ppt
intro.pptintro.ppt
intro.ppt
 

More from MediaMosa

MediaMosa as a Mass-storage Solution - 11 december 2013, Brussels
MediaMosa as a Mass-storage Solution - 11 december 2013, BrusselsMediaMosa as a Mass-storage Solution - 11 december 2013, Brussels
MediaMosa as a Mass-storage Solution - 11 december 2013, BrusselsMediaMosa
 
Drupalcafe meets MediaMosa - 17 october 2013, Utrecht
Drupalcafe meets MediaMosa  - 17 october 2013, UtrechtDrupalcafe meets MediaMosa  - 17 october 2013, Utrecht
Drupalcafe meets MediaMosa - 17 october 2013, UtrechtMediaMosa
 
To be continued... Completing the lifecycle of innovation - TNC2013, 5 june 2013
To be continued... Completing the lifecycle of innovation - TNC2013, 5 june 2013To be continued... Completing the lifecycle of innovation - TNC2013, 5 june 2013
To be continued... Completing the lifecycle of innovation - TNC2013, 5 june 2013MediaMosa
 
MediaMosa Foundation - Webstroom 15 april 2013
MediaMosa Foundation - Webstroom  15 april 2013MediaMosa Foundation - Webstroom  15 april 2013
MediaMosa Foundation - Webstroom 15 april 2013MediaMosa
 
MediaSalsa - Inuits and RUG - Video Vendor Event - 19 juni 2012 - Pitch prese...
MediaSalsa - Inuits and RUG - Video Vendor Event - 19 juni 2012 - Pitch prese...MediaSalsa - Inuits and RUG - Video Vendor Event - 19 juni 2012 - Pitch prese...
MediaSalsa - Inuits and RUG - Video Vendor Event - 19 juni 2012 - Pitch prese...MediaMosa
 
MediaSalsa - Inuits and RUG - Video Vendor Event - 19 juni 2012
MediaSalsa - Inuits and RUG - Video Vendor Event - 19 juni 2012MediaSalsa - Inuits and RUG - Video Vendor Event - 19 juni 2012
MediaSalsa - Inuits and RUG - Video Vendor Event - 19 juni 2012MediaMosa
 
Inuits en RUG - MediaSalsa - Video Vendor Event 19 juni 2012 - Short
Inuits en RUG - MediaSalsa - Video Vendor Event 19 juni 2012 - ShortInuits en RUG - MediaSalsa - Video Vendor Event 19 juni 2012 - Short
Inuits en RUG - MediaSalsa - Video Vendor Event 19 juni 2012 - ShortMediaMosa
 
A new approach for adding metadata to online Media - TNC2012
A new approach for adding metadata to online Media - TNC2012A new approach for adding metadata to online Media - TNC2012
A new approach for adding metadata to online Media - TNC2012MediaMosa
 
Videoservices via SURFconext - 29 maart 2012
Videoservices via SURFconext  - 29 maart 2012Videoservices via SURFconext  - 29 maart 2012
Videoservices via SURFconext - 29 maart 2012MediaMosa
 
MediaMosa – BasicLTI Provider - Community day - 8 december 2011
MediaMosa – BasicLTI Provider - Community day - 8 december 2011MediaMosa – BasicLTI Provider - Community day - 8 december 2011
MediaMosa – BasicLTI Provider - Community day - 8 december 2011MediaMosa
 
MediaMosa 3.x Release 
& Transcripting Project - Community day - 8 december ...
MediaMosa 3.x Release 
& Transcripting Project  - Community day - 8 december ...MediaMosa 3.x Release 
& Transcripting Project  - Community day - 8 december ...
MediaMosa 3.x Release 
& Transcripting Project - Community day - 8 december ...MediaMosa
 
MediaMosa Future - Community day 8 december 2011
MediaMosa Future - Community day 8 december 2011MediaMosa Future - Community day 8 december 2011
MediaMosa Future - Community day 8 december 2011MediaMosa
 
Rich Media Extra - MediaMosa Ingestor
Rich Media Extra - MediaMosa IngestorRich Media Extra - MediaMosa Ingestor
Rich Media Extra - MediaMosa IngestorMediaMosa
 
WCAG compliancy for MediaMosa
WCAG compliancy for MediaMosaWCAG compliancy for MediaMosa
WCAG compliancy for MediaMosaMediaMosa
 
Using OpenSource LCMS Chamilo 2.0 with MediaMosa
Using OpenSource LCMS Chamilo 2.0 with MediaMosaUsing OpenSource LCMS Chamilo 2.0 with MediaMosa
Using OpenSource LCMS Chamilo 2.0 with MediaMosaMediaMosa
 
Media & Learning Brussels 24 November 2011
Media & Learning Brussels 24 November 2011Media & Learning Brussels 24 November 2011
Media & Learning Brussels 24 November 2011MediaMosa
 
MediaMosa Player v2_0 - OPEN SOURCE HTML5 WEBLECTURES PLAYER
MediaMosa Player v2_0 - OPEN SOURCE HTML5 WEBLECTURES PLAYERMediaMosa Player v2_0 - OPEN SOURCE HTML5 WEBLECTURES PLAYER
MediaMosa Player v2_0 - OPEN SOURCE HTML5 WEBLECTURES PLAYERMediaMosa
 
Seminar 20111122 - MediaMosa projects
Seminar 20111122 - MediaMosa projectsSeminar 20111122 - MediaMosa projects
Seminar 20111122 - MediaMosa projectsMediaMosa
 
Norwegian studietur MediaMosa 22 november 2011
Norwegian studietur MediaMosa   22 november 2011Norwegian studietur MediaMosa   22 november 2011
Norwegian studietur MediaMosa 22 november 2011MediaMosa
 
Rondetafel bijeenkomst MediaMosa -17 november 2011
Rondetafel bijeenkomst MediaMosa -17 november 2011Rondetafel bijeenkomst MediaMosa -17 november 2011
Rondetafel bijeenkomst MediaMosa -17 november 2011MediaMosa
 

More from MediaMosa (20)

MediaMosa as a Mass-storage Solution - 11 december 2013, Brussels
MediaMosa as a Mass-storage Solution - 11 december 2013, BrusselsMediaMosa as a Mass-storage Solution - 11 december 2013, Brussels
MediaMosa as a Mass-storage Solution - 11 december 2013, Brussels
 
Drupalcafe meets MediaMosa - 17 october 2013, Utrecht
Drupalcafe meets MediaMosa  - 17 october 2013, UtrechtDrupalcafe meets MediaMosa  - 17 october 2013, Utrecht
Drupalcafe meets MediaMosa - 17 october 2013, Utrecht
 
To be continued... Completing the lifecycle of innovation - TNC2013, 5 june 2013
To be continued... Completing the lifecycle of innovation - TNC2013, 5 june 2013To be continued... Completing the lifecycle of innovation - TNC2013, 5 june 2013
To be continued... Completing the lifecycle of innovation - TNC2013, 5 june 2013
 
MediaMosa Foundation - Webstroom 15 april 2013
MediaMosa Foundation - Webstroom  15 april 2013MediaMosa Foundation - Webstroom  15 april 2013
MediaMosa Foundation - Webstroom 15 april 2013
 
MediaSalsa - Inuits and RUG - Video Vendor Event - 19 juni 2012 - Pitch prese...
MediaSalsa - Inuits and RUG - Video Vendor Event - 19 juni 2012 - Pitch prese...MediaSalsa - Inuits and RUG - Video Vendor Event - 19 juni 2012 - Pitch prese...
MediaSalsa - Inuits and RUG - Video Vendor Event - 19 juni 2012 - Pitch prese...
 
MediaSalsa - Inuits and RUG - Video Vendor Event - 19 juni 2012
MediaSalsa - Inuits and RUG - Video Vendor Event - 19 juni 2012MediaSalsa - Inuits and RUG - Video Vendor Event - 19 juni 2012
MediaSalsa - Inuits and RUG - Video Vendor Event - 19 juni 2012
 
Inuits en RUG - MediaSalsa - Video Vendor Event 19 juni 2012 - Short
Inuits en RUG - MediaSalsa - Video Vendor Event 19 juni 2012 - ShortInuits en RUG - MediaSalsa - Video Vendor Event 19 juni 2012 - Short
Inuits en RUG - MediaSalsa - Video Vendor Event 19 juni 2012 - Short
 
A new approach for adding metadata to online Media - TNC2012
A new approach for adding metadata to online Media - TNC2012A new approach for adding metadata to online Media - TNC2012
A new approach for adding metadata to online Media - TNC2012
 
Videoservices via SURFconext - 29 maart 2012
Videoservices via SURFconext  - 29 maart 2012Videoservices via SURFconext  - 29 maart 2012
Videoservices via SURFconext - 29 maart 2012
 
MediaMosa – BasicLTI Provider - Community day - 8 december 2011
MediaMosa – BasicLTI Provider - Community day - 8 december 2011MediaMosa – BasicLTI Provider - Community day - 8 december 2011
MediaMosa – BasicLTI Provider - Community day - 8 december 2011
 
MediaMosa 3.x Release 
& Transcripting Project - Community day - 8 december ...
MediaMosa 3.x Release 
& Transcripting Project  - Community day - 8 december ...MediaMosa 3.x Release 
& Transcripting Project  - Community day - 8 december ...
MediaMosa 3.x Release 
& Transcripting Project - Community day - 8 december ...
 
MediaMosa Future - Community day 8 december 2011
MediaMosa Future - Community day 8 december 2011MediaMosa Future - Community day 8 december 2011
MediaMosa Future - Community day 8 december 2011
 
Rich Media Extra - MediaMosa Ingestor
Rich Media Extra - MediaMosa IngestorRich Media Extra - MediaMosa Ingestor
Rich Media Extra - MediaMosa Ingestor
 
WCAG compliancy for MediaMosa
WCAG compliancy for MediaMosaWCAG compliancy for MediaMosa
WCAG compliancy for MediaMosa
 
Using OpenSource LCMS Chamilo 2.0 with MediaMosa
Using OpenSource LCMS Chamilo 2.0 with MediaMosaUsing OpenSource LCMS Chamilo 2.0 with MediaMosa
Using OpenSource LCMS Chamilo 2.0 with MediaMosa
 
Media & Learning Brussels 24 November 2011
Media & Learning Brussels 24 November 2011Media & Learning Brussels 24 November 2011
Media & Learning Brussels 24 November 2011
 
MediaMosa Player v2_0 - OPEN SOURCE HTML5 WEBLECTURES PLAYER
MediaMosa Player v2_0 - OPEN SOURCE HTML5 WEBLECTURES PLAYERMediaMosa Player v2_0 - OPEN SOURCE HTML5 WEBLECTURES PLAYER
MediaMosa Player v2_0 - OPEN SOURCE HTML5 WEBLECTURES PLAYER
 
Seminar 20111122 - MediaMosa projects
Seminar 20111122 - MediaMosa projectsSeminar 20111122 - MediaMosa projects
Seminar 20111122 - MediaMosa projects
 
Norwegian studietur MediaMosa 22 november 2011
Norwegian studietur MediaMosa   22 november 2011Norwegian studietur MediaMosa   22 november 2011
Norwegian studietur MediaMosa 22 november 2011
 
Rondetafel bijeenkomst MediaMosa -17 november 2011
Rondetafel bijeenkomst MediaMosa -17 november 2011Rondetafel bijeenkomst MediaMosa -17 november 2011
Rondetafel bijeenkomst MediaMosa -17 november 2011
 

MediaMosa Transcription technology

  • 1. Speech & Language Technology A.J.vanHessen@ewi.utwente.nl   Searching  in  spoken  words Disclosure of recorded content in MediaMosa SURFnet Relatiedagen 2010 Noordwijkerhout December 9, 2010
  • 2. •  Introduction   –  Why  speech  is  so  important   –  What  is  HLT?   •  Working  applications:   –  Self-­‐service  (Internet    Telephony)   –  Searching  in  recorded  audiovisual  recordings   •  Demonstrations  
  • 3. Humans as speaking creatures The  start  of  the  human   speech  started  some   100.000  years  ago.   Before,  the  shape  of  the   vocal  track  was  not  “ready”   for  the  modern  speech.  The   larynx  was  situated  too   high,  something  you  can  see   with  chimps.  
  • 4. Humans as writing creatures Sumer  (3300  AD,  Mesopotamia)  is   probably  the  oldest  written  language.   NU   -­‐100.000   -­‐10.000   -­‐3300   speech   farming   text  
  • 5. What  is  HLT?   •  Human  Language  Technology  is  the   technology  that  mimics  the  human  language   capacity.   speech text sign
  • 6. Redundancy   •  Vlgones  een  oznrdeeok  op  een  Eglnese   uvinretsiet  mkaat  het  neit  uitin  wlkee   vloogdre  de  ltteers  in  een  wrood  saatn,  het   einge  watblegnaijrk  is  is  dat  de  eretse  en   de  ltaatse  ltteer  op  de  jiutsepatals  saatn.   De  rset  van  de  ltteers  mgoen  wllikueirg   gpletaastwdoren  en  je  knut  vrelvogens   gwoeon  lzeen  wat  er  saatt.  Dit  kmotodmat   we  neit  ekle  ltteer  op  zcih  lzeen  maar  het   wrood  als  gheeel.  
  • 7. Pensez a ceque v s avez f nit et dema n dez vs i est coq ue 3 ai
  • 8. Dialogue  systems  (telephony,  real  time,  limited  complexity)   Disclosure  systems  (high  quality  audio,  ofQline,  complex)   WORKING  APPLICATIONS  
  • 9. Spoken  Document   ContactCenter   Retrieval Information   Retrieval Voice   HLT Natural  Language   Search   Web   Mens-­‐Machine-­‐ Emotie  detectie: Mobile   Dictation Lachen/Huilen Communicatie
  • 10. Companies  using  speech  technology  
  • 11. How  may  I  help  you   Why are they Who is calling? calling? Classification based on Identification via ZIP- the recognition of the code and house number question: “how may I help you”
  • 12. OrganisaCons  using     speech  technology  
  • 13. Disclosure  of  audiovisual  archives   •  The  number  of  AV-­‐archives  on  the  Internet   increases  rapidly   •  Archiving  is  not  enough:  disclosure  and   reusing  is  required!   •  The  use  of  HLT  is  needed  (humans  cost  too   much).  
  • 14. Digitalized (historic) LVSR   Digital recorded collections collections WFH H.M. Koningin Second feministic wave Wilhelmina Buchenwald Memories of Indonesia
  • 15. Searching  in  historic  radio  recordings:   Radio  Oranje  
  • 17. Oral  History:  Brandgrens,  RoKerdam   10  getuigen  van  het  bombardement  van  Rotterdam  (mei  ‘40)  vertellen  hun   verhaal.  TST  wordt  gebruikt  om  in  de  getuigenissen  te  zoeken.  
  • 18. Searching  in  the  radio  interviews  of  WFH  
  • 19. Searching  in  46  interview  collecCons:   getuigenverhalen  (600  hour)  
  • 20. Searching  in  500  interviews  in  CroaCa  
  • 21. CroMe  -­‐  Audio  Search   Searching  for:   commandant   Phrase   boundaries   Found  word   5  fragments   (5x  commandant)   found  
  • 22. CroMe  -­‐  Audio  Search   Search  word   traumas   Language   found  
  • 24. Parliament   transcriptions    Gisteren    was    er    een    bespreking        ivm        de    betrekkingen  tussen  Nederland  en  Vlaanderen  
  • 25. RecogniCon  of  lectures   •  Record  the  speech   •  Record  the  PPT   •  Recognise  the  speech   •  Use  the  display  time  of  each  slide  as  THE   time  unit   •  Use  the  recognised  speech  as  keywords  for   each  slide  
  • 26. Searching  in  news  broadcasts  
  • 27. Metadata  -­‐  Language  model   Text in the slide(s) Lectures handouts Language model Environmental texts