Mulist

Multi Lingual Speech translator

Transcript

1. MuLiST  :  Mul*-­‐Lingual   Speech  Translator   Batch Members: Apinaya Prethi K N (070105122005) Dhilip C (070105122015) Pavithra R (070105122066) 25-­‐Sep-­‐12   Department  Of  Computer  Science  and  Engineering   1   Project Guide: Mr. Prabhakar M.E Lecturer CSE Batch no : 22 Review no : Zero
2. Objec*ve    To  design  a  speech-­‐to-­‐speech  translator,   which  can  translate  speech  from  any  language   into  English  Speech.  MuLiST  (Mul*-­‐Lingual   Speech  Translator)  can  also  be  configured  by  the   end-­‐user  to  recognize  and  translate  new   languages.     25-­‐Sep-­‐12   Department  Of  Computer  Science  and   Engineering   2  
3. Abstract    MuLiST,  a  Spoken  Language  Translator   (SLT)  is  a  module  that  recognises  the  speech  of   source  language,  compares  it  with  the   templates,  converts  it  into  templates  of  the   des*na*on  language,  synthesizes  the  target   language  and  speaks  it.   25-­‐Sep-­‐12   Department  Of  Computer  Science  and   Engineering   3  
4. Introduc*on   •  Transla*on  is  the  communica*on  of  the   meaning  of  a  source-­‐language  by  means  of  an   equivalent  target-­‐language.   •  A  Spoken  Language  Translator  (SLT)  is  a   module  that  encapsulates  both  Natural   Language  Processing  (NLP)  and  a  Translator.   25-­‐Sep-­‐12   Department  Of  Computer  Science  and   Engineering   4  
5. Exis*ng  Features   •  Language  defini*on  can  only  be  configured  by   the  programmers  and  is  not  configurable  by     the  user.   •  Only  the  developers  can  add  any  new   languages,  which  is  usually  a  complicated   process.   •  Custom  language  defini*ons,  templates  and   dic*onaries  cannot  be  shared  with  other   users.   25-­‐Sep-­‐12   Department  Of  Computer  Science  and   Engineering   5  
6. Proposed  Features   •  To  build  a  user  configurable  language   defini*on  system,  were  an  end-­‐user  can  also   add  a  new  language.   •  To  incorporate  learning  element  in  the  system   to  automa*cally  adopt  according  to  users.   •  To  build  a  new  feature  that  enables  users  to   share  their  custom  language  defini*on,   templates  and  dic*onaries  with  other  users  of   the  system.   25-­‐Sep-­‐12   Department  Of  Computer  Science  and   Engineering   6  
7. Algorithm  and  Methodology   Sound  sound  =  new  Sound()   Template  template  =  Template(sound)   While(rw  !  =  null)   {    RecoWords  rw  =  RecoWords(template)    Words  words  =  +Words(rw)   }   Synthesize(words)     25-­‐Sep-­‐12   Department  Of  Computer  Science  and   Engineering   7  
8. Conclusion    This  paper  provides  the  methodologies   and  tools  for  the  transla*on  of  source  language   into  target  language  and  thereby  allow  users  to   share  and  customize  their  language  libraries.  It   also  allows  end  users  to  add  and  configure  a   new  language.   25-­‐Sep-­‐12   Department  Of  Computer  Science  and   Engineering   8  
9. References   •  Bowen Zhou, Yuqing Gao, Liang Gu,“IBM MASTOR SYSTEM: Multilingual Automatic Speech-to-speech Translator.” •  Gao Y , J. Sorensen, H. B. Zhou Z. Diao, “A Trainable Approach for Multi-Lingual Speech-To-Speech Translation System” •  Ruiqiang Zhang. Genichiro Kikui and Hirofumi Yamamoto, “A Unified Approach in Speech-to-Speech Translation: Integrating Features of Speech Recognition and Machine Translation” 25-­‐Sep-­‐12   Department  Of  Computer  Science  and   Engineering   9  
10. Contd.   •  C#  4  and  .NET  4,  First  Edi*on,  Wrox   professional.   •  hbp://msdn.microsod.com/en-­‐us/library/ ms723627%28v=vs.85%29.aspx   •  hbp://msdn.microsod.com/en-­‐us/default   25-­‐Sep-­‐12   Department  Of  Computer  Science  and   Engineering   10