Adapting language models with a bit mask for a subset of related words

• US 8,589,163 B2
• Filed: 12/04/2009
• Issued: 11/19/2013
• Est. Priority Date: 12/04/2009
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: AT&T Intellectual Property I LP (AT&T, Inc.)
• Inventors: Ljolje, Andrej, Gilbert, Mazin
• Primary Examiner(s): Lerner, Martin
• Application Number:
  US12/631,111
• Publication Number:
  US 20110137653A1
• Time in Patent Office: 1,446 Days
• Field of Search: 704/235, 704/236, 704/243, 704/244, 704/251, 704/255, 704/252, 704/257, 704/242
• US Class Current: 704/255
• CPC Class Codes: G10L 15/183 using context dependencies, e.g. language models
• CPC Class Codes: G10L 2015/227 of the speaker; Human-factor methodology
• CPC Class Codes: G10L 2015/228 of application context

Abstract

Disclosed herein are systems, methods, and computer-readable storage media for performing speech recognition based on a masked language model. A system configured to practice the method receives a masked language model including a plurality of words, wherein a bit mask identifies whether each of the plurality of words is allowed or disallowed with regard to an adaptation subset, receives input speech, generates a speech recognition lattice based on the received input speech using the masked language model, removes from the generated lattice words identified as disallowed by the bit mask for the adaptation subset, and recognizes the received speech based on the lattice. Alternatively during the generation step, the system can only add words indicated as allowed by the bit mask. The bit mask can be separate from or incorporated as part of the masked language model. The system can dynamically update the adaptation subset and bit mask.

Claims

1. A computer-readable storage device having instructions stored which, when executed by a processor, cause the processor to perform operations comprising:

receiving, via the processor, a language model having a vocabulary of words, the language model comprising a speech recognizer output lattice;

receiving a user query;

identifying, via the processor, an adaptation subset of related words in the vocabulary of words in the language model based on the user query;

adding a geographically-specific bit mask to each word in the vocabulary of words based on the adaptation subset, thereby generating a masked language model; and

modifying weights of other words in the speech recognizer output lattice based on the masked language model.

2. The computer-readable storage device of claim 1, the computer-readable storage device having additional instructions stored which result in the operations further comprising:

identifying a plurality of adaptation subsets based on the user query; and

adding a separate bit mask to each word in the vocabulary of words corresponding to each of the plurality of adaptation subsets.

3. The computer-readable storage device of claim 2, wherein the plurality of adaptation subsets is a multiple of a power of two.

4. The computer-readable storage device of claim 1, wherein the identifying of the adaptation subset of related words is based on a geographical relationship.

5. The computer-readable storage device of claim 1, wherein the identifying of the adaptation subset of related words is based on one of a speaker usage pattern and a business category.

6. The computer-readable storage device of claim 1, wherein a speech recognition application allows only words in the language model having a matching bit mask value for a given adaptation subset.

7. The computer-readable storage device of claim 1, wherein the geographically-specific bit mask comprises, for each word in the vocabulary, one of a 1 and a 0.

8. A system comprising:

a processor; and

a computer-readable storage medium having instructions stored which, when executed by the processor, cause the processor to perform operations comprising;

receiving, via the processor, a language model having a vocabulary of words, the language model comprising a speech recognizer output lattice;

receiving a user query;

identifying, via the processor, an adaptation subset of related words in the vocabulary of words in the language model based on the user query;

adding a geographically-specific bit mask to each word in the vocabulary of words based on the adaptation subset, thereby generating a masked language model; and

modifying weights of other words in the speech recognizer output lattice based on the masked language model.

9. The system of claim 8, the computer-readable storage medium having additional instructions stored which result in the operations further comprising:

identifying a plurality of adaption subsets based on the user query; and

adding a separate dynamically generated bit mask to each word in the vocabulary of words corresponding to each of the plurality of adaption subsets.

10. The system of claim 8, wherein the geographically-specific bit mask is separate from, but associated with, the masked language model.

11. The system of claim 8, the computer-readable storage medium having additional instructions stored which result in the operations further comprising updating the adaptation subset and the geographically-specific bit mask based on a usage history.

12. The system of claim 8, wherein the adaptation subset and the geographically-specific bit mask are associated with a user profile.

13. The system of claim 8, the computer-readable storage medium having additional instructions stored which result in the operations further comprising enabling and disabling the geographically-specific bit mask.

14. The system of claim 8, wherein the geographically-specific bit mask comprises, for each of the plurality of words, one of a 1 and a 0.

15. A method comprising:

receiving, via the processor, a language model having a vocabulary of words, the language model comprising a speech recognizer output lattice;

receiving a user query;

identifying, via the processor, an adaptation subset of related words in the vocabulary of words in the language model based on the user query;

adding a geographically-specific bit mask to each word in the vocabulary of words based on the adaptation subset, thereby generating a masked language model; and

modifying weights of other words in the speech recognizer output lattice based on the masked language model.

16. The method of claim 15, further comprising:

identifying a plurality of adaptation subsets based on the user query; and

adding a separate bit mask to each word in the vocabulary of words corresponding to each of the plurality of adaptation subsets.

17. The method of claim 15, the method further comprising dynamically generating the adaptation subset and the geographically-specific bit mask.

18. The method of claim 15, wherein identifying the adaptation subset of related words is based on a geographical relationship.

19. The method of claim 15, wherein a speech recognition application allows only words in the language model having a matching bit mask value for a given adaptation subset.

20. The method of claim 15, wherein a speech recognition application allows only words in the language model having a matching bit mask value for each of a plurality of given adaptation subsets.