Speech-to Speech Translation       


  • A Real-time and Mobile Solution to Mitigate Language Barriers

  • Audibly and visually translates between two languages
  • 2-way translation of “free form” conversational speech
  • Target for instantaneous & highly accurate Speech-to-Speech (S2S) translation on mobile devices (handhelds, smartphone etc.)
  • No need for server connectivity
  • Robust & speaker independent large vocabulary speech recognition, accommodating differences in tone and accent with online speaker adaptation
  • Data-driven statistical machine translation Machine learning techniques ubiquitously applied, which enables rapid development for new languages & domains

Video Demo
http://researcher.ibm.com/view_project_subpage.php?id=2286


Recent Publications
http://researcher.ibm.com/view_project_subpage.php?id=2325



Vocabularies

English-Chinese English-Arabic

Vocabulary size

– 50,000 unique English words

– 45,000 unique Chinese words

Vocabulary size

– 50,000 unique English words

– 200,000 unique Arabic words

Designed Domains

English-Chinese English-Arabic

  • Travel & Tourism
  • Airport (Customs, Duty free goods, Visa, Money Transfer)
  • Hotel reservation– Dining, Food and Restaurant
  • Culture and Entertainment
  • Traffic and Directions
  • Sightseeing & Shopping

  • Business (Rental service, Advertisement, Tax, Import and Export)
  • Job Interview– Introduction & greeting
  • Health and Medical Care (Diagnose, Pharmacy)
  • Sports (Balls, Aquatics, Track... Olympic Games and Venues)
  • Auto and Ground Transportation
  • and more…


For more information, please contact:

Bowen Zhou, PhD
Manager & Researcher
Speech-to-speech Translation Group
Email: last name AT us.ibm.com