×

Multi-lingual speech recognition with cross-language context modeling

DC
  • US 7,149,688 B2
  • Filed: 11/04/2002
  • Issued: 12/12/2006
  • Est. Priority Date: 11/04/2002
  • Status: Active Grant
First Claim
Patent Images

1. A method for selecting context-dependent units for speech recognition comprising:

  • accepting a representation of the first word from a first language in terms of subword units from a first set of subword units that is associated with the first language, including accepting a first subword unit of the first word which is adjacent to the second word from a second language according to a speech recognition grammar;

    determining a common set of features that characterize subword units in both the first set of subword units and in a second set of subword units that is associated with the second language; and

    selecting one or more first context-dependent units corresponding to the first subword unit according to features in the common set of features of a second subword unit of the second word which is adjacent to the first word according to the grammar.

View all claims
  • 6 Assignments
Timeline View
Assignment View
    ×
    ×