Method, apparatus and computer program product for providing adaptive language model scaling

• US 7,716,049 B2
• Filed: 06/30/2006
• Issued: 05/11/2010
• Est. Priority Date: 06/30/2006
• Status: Expired due to Fees

• Current Assignee: Nokia Corporation
• Original Assignee: Nokia Corporation
• Inventors: Tian, Jilei
• Primary Examiner(s): Sked; Matthew J
• Application Number:
  US11/427,954
• Publication Number:
  US 20080004877A1
• Time in Patent Office: 1,411 Days
• Field of Search: None
• US Class Current: 704/251
• CPC Class Codes: G10L 15/197 Probabilistic grammars, e.g. word n-grams

Abstract

An apparatus for providing adaptive language model scaling includes an adaptive scaling element and an interface element. The adaptive scaling element is configured to receive input speech comprising a sequence of spoken words and to determine a plurality of candidate sequences of text words in which each of the candidate sequences has a corresponding sentence score representing a probability that a candidate sequence matches the sequence of spoken words. Each corresponding sentence score is calculated using an adaptive scaling factor. The interface element is configured to receive a user input selecting one of the candidate sequences. The adaptive scaling element is further configured to estimate an objective function based on the user input and to modify the adaptive scaling factor based on the estimated objective function.

Claims

1. A method comprising:

receiving input speech comprising a sequence of spoken words;

determining, via a processor, a plurality of candidate sequences of text words in which each of the candidate sequences has a corresponding sentence score representing a probability that a candidate sequence matches the sequence of spoken words, each corresponding sentence score being calculated using an adaptive scaling factor that applies weights to respective components of word scores used to determine the corresponding sentence score;

receiving a user input selecting one of the candidate sequences;

estimating an objective function based on the user input, the objective function defining a difference value to be applied to the adaptive scaling factor based on the user input to improve a sentence score of the user selected one of the candidate sequences; and

modifying the adaptive scaling factor by adjusting the adaptive scaling factor by an amount determined by the estimated objective function.

2. A method according to claim 1, further comprising generating a word lattice including a series of candidate words corresponding to each of the spoken words in the sequence of spoken words, each of the candidate words having a corresponding word score.

3. A method according to claim 2, further comprising determining the word score of each of the candidate words by summing an acoustic score and a language score for each word, wherein at least one of the acoustic score or the language score is weighted according to the adaptive scaling factor.

4. A method according to claim 3, wherein determining the plurality of candidate sequences of text words comprises determining each of the candidate sequences responsive to a summation of selected word candidates.

5. A method according to claim 1, further comprising calculating the sentence score by summing an acoustic score and a language score for each candidate sequence, and wherein at least one of the acoustic score or the language score is weighted according to the adaptive scaling factor.

6. A method according to claim 1, wherein estimating the objective function is performed based on the user input and a collection of recently used candidate sequences.

7. A method according to claim 6, further comprising applying a forgetting factor to define the collection of recently used candidate sequences.

8. A method according to claim 7, wherein applying the forgetting factor comprises one of:

defining the collection based on age of recently used sentences;

or

defining the collection to be a predetermined number of recently used sentences.

9. A computer program product comprising at least one computer-readable storage medium having computer-readable program code portions stored therein, the computer-readable program code portions comprising:

a first executable portion for receiving input speech comprising a sequence of spoken words;

a second executable portion for determining a plurality of candidate sequences of text words in which each of the candidate sequences has a corresponding sentence score representing a probability that a candidate sequence matches the sequence of spoken words, each corresponding sentence score being calculated using an adaptive scaling factor that applies weights to respective components of word scores used to determine the corresponding sentence score;

a third executable portion for receiving a user input selecting one of the candidate sequences;

a fourth executable portion for estimating an objective function based on the user input, the objective function defining a difference value to be applied to the adaptive scaling factor based on the user input to improve a sentence score of the user selected one of the candidate sequences; and

a fifth executable portion for modifying the adaptive scaling factor by adjusting the adaptive scaling factor by an amount determined by the estimated objective function.

10. A computer program product according to claim 9, further comprising a sixth executable portion for generating a word lattice including a series of candidate words corresponding to each of the spoken words in the sequence of spoken words, each of the candidate words having a corresponding word score.

11. A computer program product according to claim 10, further comprising a seventh executable portion for determining the word score of each of the candidate words by summing an acoustic score and a language score for each word, wherein at least one of the acoustic score or the language score is weighted according to the adaptive scaling factor.

12. A computer program product according to claim 11, wherein the second executable portion includes instructions for determining each of the candidate sequences responsive to a summation of selected word candidates.

13. A computer program product according to claim 9, further comprising a sixth executable portion for calculating the sentence score by summing an acoustic score and a language score for each candidate sequence, and wherein at least one of the acoustic score or the language score is weighted according to the adaptive scaling factor.

14. A computer program product according to claim 9, wherein the fourth executable portion is executed based on the user input and a collection of recently used candidate sequences.

15. A computer program product according to claim 14, further comprising a sixth executable portion for applying a forgetting factor to define the collection of recently used candidate sequences.

16. A computer program product according to claim 15, wherein the sixth executable portion includes instructions for one of:

defining the collection based on age of recently used sentences;

or

defining the collection to be a predetermined number of recently used sentences.

17. An apparatus comprising a processor and memory including computer program code, the memory and the computer program code configured to, with the processor, cause the apparatus to at least:

receive input speech comprising a sequence of spoken words;

determine a plurality of candidate sequences of text words in which each of the candidate sequences has a corresponding sentence score representing a probability that a candidate sequence matches the sequence of spoken words, each corresponding sentence score being calculated using an adaptive scaling factor that applies weights to respective components of word scores used to determine the corresponding sentence score;

receive a user input selecting one of the candidate sequences,

estimate an objective function based on the user input, the objective function defining a difference value to be applied to the adaptive scaling factor based on the user input to improve a sentence score of the user selected one of the candidate sequences; and

modify the adaptive scaling factor by adjusting the adaptive scaling factor by an amount determined by the estimated objective function.

18. An apparatus according to claim 17, wherein the memory and the computer program code are configured to, with the processor, cause the apparatus to generate a word lattice including a series of candidate words corresponding to each of the spoken words in the sequence of spoken words, each of the candidate words having a corresponding word score.

19. An apparatus according to claim 18, wherein the memory and the computer program code are configured to, with the processor, cause the apparatus to determine the word score of each of the candidate words by summing an acoustic score and a language score for each word, wherein at least one of the acoustic score or the language score is weighted according to the adaptive scaling factor.

20. An apparatus according to claim 19, wherein the adaptive memory and the computer program code are configured to, with the processor, cause the apparatus to determine each of the candidate sequences responsive to a summation of selected word candidates.

21. An apparatus according to claim 17, wherein the memory and the computer program code are configured to, with the processor, cause the apparatus to calculate the sentence score by summing an acoustic score and a language score for each candidate sequence, and wherein at least one of the acoustic score or the language score is weighted according to the adaptive scaling factor.

22. An apparatus according to claim 17, wherein the memory and the computer program code are configured to, with the processor, cause the apparatus to estimate the objective function based on the user input and a collection of recently used candidate sequences.

23. An apparatus according to claim 22, wherein the memory and the computer program code are configured to, with the processor, cause the apparatus to apply a forgetting factor to define the collection of recently used candidate sequences.

24. An apparatus according to claim 23, wherein the memory and the computer program code are configured to, with the processor, cause the apparatus to perform one of:

defining the collection based on age of recently used sentences;

or

defining the collection to be a predetermined number of recently used sentences.

25. An apparatus according to claim 17, wherein the apparatus is embodied as a mobile terminal.

26. An apparatus comprising:

means for receiving input speech comprising a sequence of spoken words;

means for determining a plurality of candidate sequences of text words in which each of the candidate sequences has a corresponding sentence score representing a probability that a candidate sequence matches the sequence of spoken words, each corresponding sentence score being calculated using an adaptive scaling factor that applies weights to respective components of word scores used to determine the corresponding sentence score;

means for receiving a user input selecting one of the candidate sequences;

means for estimating an objective function based on the user input, the objective function defining a difference value to be applied to the adaptive scaling factor based on the user input to improve a sentence score of the user selected one of the candidate sequences; and

means for modifying the adaptive scaling factor by adjusting the adaptive scaling factor by an amount determined by the estimated objective function.