Homonym processing in the context of voice-activated command systems

• US 20060004572A1
• Filed: 09/07/2004
• Published: 01/05/2006
• Est. Priority Date: 06/30/2004
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Ju, Yun-Cheng, Ollason, David G., Bhatia, Siddharth
• Granted Patent: US 7,181,387 B2
• Time in Patent Office: Days
• US Class Current: 704/243
• CPC Class Codes: G10L 15/06 Creation of reference templates; Training of speech recognition systems, e.g. adaptation to the characteristics of the speaker's voice G10L15/14 takes precedence
• CPC Class Codes: G10L 15/187 Phonemic context, e.g. pronunciation rules, phonotactical constraints or phoneme n-grams

Abstract

A method is disclosed from constructing a grammar. The grammar is configured to be processed by a speech recognition engine in the context of a voice-activated command system. The method includes receiving a database containing a plurality of terms. From the plurality of terms, first and second terms are identified. The first and second terms are spelled differently but have a first pronunciation in common. One of the first and second terms also has a second pronunciation that is not inherent to the other of the first and second terms. The first and second pronunciations are placed within the grammar.

Claims

1-20. -20. (canceled)

21. A method for constructing a grammar to be processed by a speech recognition engine in the context of a voice-activated command system, the method comprising:

receiving a database containing a plurality of terms;

identifying from said plurality a first term and a second term that are spelled differently but have a first pronunciation in common, wherein one of the first and second terms also has a second pronunciation that is not inherent to the other of the first and second terms; and

placing the first and second pronunciations within the grammar.

22. The method of claim 21, wherein identifying a first term and a second term comprises obtaining a pronunciation for each of the plurality of terms in the database.

23. The method of claim 22, wherein obtaining a pronunciation comprises obtaining a pronunciation from a speech recognition dictionary.

24. The method of claim 22, wherein obtaining a pronunciation comprises obtaining a pronunciation from an application dictionary.

25. The method of claim 22, wherein identifying a first and second term comprises organizing the plurality of terms into a plurality of pronunciation classes, wherein each pronunciation class corresponds to a distinct pronunciation, wherein both of the first and second terms are included in a same pronunciation class, and wherein one of the first and second terms is included in a class that the other is not in.

26. The method of claim 22, wherein identifying a first term and a second term comprises identifying a first term and second term having quasi-homonym characteristics.

27. A computer-implemented method for accomplishing disambiguation in the context of a voice-dialing system, the method comprising:

providing an input to a speech recognition engine for processing relative to a grammar that corresponds to a database containing a plurality of terms;

including in the grammar a pair of pronunciations that correspond to a pair of terms from said plurality that are spelled differently, one of said pronunciations being a pronunciation shared by the pair of terms and the other being unique to one of the pair of terms;

receiving from the speech recognition engine an output corresponding to one of said pair of pronunciations;

utilizing, based at least in part on the output, either both or one of the pair of terms as a basis for disambiguation.

28. The method of claim 27, wherein utilizing comprises utilizing both of the pair of terms when the output corresponds to a pronunciation shared by the pair.

29. The method of claim 27, wherein utilizing comprises utilizing one of the pair of terms when the output corresponds to one that is unique to one of the pair of terms.

30. The method of claim 27, wherein providing an input to a speech recognition engine comprises receiving a speech input and providing a representation of the speech input to the speech recognition engine.

31. The method of claim 30, wherein receiving a speech input comprises receiving a speech input from a caller.

32. The method of claim 31, wherein receiving a speech input comprises receiving a spoken name.

33. The method of claim 27, wherein receiving an output comprises receiving a textual representation of a person'"'"'s name.

34. The method of claim 27, wherein utilizing both of the pair of terms as a basis for disambiguation comprises:

presenting a spelling of each of the pair of terms to a caller;

receiving an input from the caller corresponding to one of the pair of terms; and

selecting for subsequent processing the one of the pair of terms corresponding to the input.

35. A context free grammar configured to be processed by a speech recognition engine in the context of a voice-activated command system, comprising:

a representation of a plurality of database terms including a pair of pronunciations that correspond to a pair of terms from said plurality that are spelled differently, one of said pronunciations being a pronunciation shared by the pair of terms and the other being unique to one of the pair of terms.