Segmenting words using scaled probabilities
Segmenting words using scaled probabilities
 CN 102,057,370 A
 Filed: 04/09/2009
 Published: 05/11/2011
 Est. Priority Date: 04/16/2008
 Status: Active Application
First Claim
Patent Images
1. method comprises:
 Receive the probability of ngram sign speech;
Determine the number of the atomic unit among the corresponding ngram;
Described number according to the atomic unit among the described ngram identifies the convergentdivergent weight;
AndThe described probability that described convergentdivergent weight is applied to described ngram sign speech determines that described ngram identifies the probability through convergentdivergent of speech.
Chinese PRB Reexamination
Abstract
Systems, methods, and apparatuses including computer program products for segmenting words using scaled probabilities. In one implementation, a method is provided. The method includes receiving a probability of a n gram identifying a word, determining a number of atomic units in the corresponding n gram, identifying a scaling weight depending on the number of atomic units in the ngram, and applying the scaling weight to the probability of the ngram identifying a word to determine a scaled probability of the ngram identifying a word.
25 Claims

1. method comprises:

Receive the probability of ngram sign speech; Determine the number of the atomic unit among the corresponding ngram; Described number according to the atomic unit among the described ngram identifies the convergentdivergent weight;
AndThe described probability that described convergentdivergent weight is applied to described ngram sign speech determines that described ngram identifies the probability through convergentdivergent of speech.


2. the method for claim 1, the described probability through convergentdivergent of wherein said ngram sign speech depends on the described number of the atomic unit among the described ngram.

3. the method for claim 1, the described probability through convergentdivergent of wherein said ngram sign speech is x ^{n}, wherein x is the described probability of described ngram sign speech, and n is the described number of the atomic unit among the described ngram.

4. the method for claim 1, the described probability through convergentdivergent of wherein said ngram sign speech is x ^{1+k (n1)}, wherein x is the described probability of described ngram sign speech, n is the described number of the atomic unit among the described ngram, and k is constant and 0≤
 k≤
1.
 k≤

5. the method for claim 1 further comprises:

Receive a plurality of symbols;
AndUse described probability that described a plurality of symbol segmentation are become speech through convergentdivergent.


6. the method for claim 1 further comprises:

Identify the ngram of less level, the ngram of described less level obtains from described ngram; The probability of each the corresponding sign speech among the ngram of reception and described less level; The described probability that the combination of the described probability of described ngram sign speech and the ngram of described less level is identified speech compares;
AndWhen the described probability of the probability of the combination sign speech of the ngram of less level and described ngram sign speech differs the assign thresholds amount, revise the corresponding described convergentdivergent weight of described probability that identifies speech with described ngram.


7. the method for claim 1 further comprises:

Receive the probability through convergentdivergent of ngram sign speech; Determine the probability through convergentdivergent of the ngram sign speech of less level, the ngram of described less level obtains from described ngram;
AndWhen the described probability through convergentdivergent that identifies speech through the probability of convergentdivergent and described ngram of the combination sign speech of the ngram of less level differs the assign thresholds amount, remove described ngram from dictionary.


8. system comprises:

Dictionary, described dictionary comprise the corresponding probability of ngram and each ngram sign speech; Zooming engine, described zooming engine comprise the convergentdivergent weight corresponding to each ngram, and described convergentdivergent weight depends on the number of the atomic unit in each ngram;
AndThe probability through convergentdivergent of each ngram sign speech determines that wherein the described probability through convergentdivergent of each ngram sign speech comprises the corresponding probability that the convergentdivergent weight is applied to each ngram sign speech.


9. system as claimed in claim 8, the described probability through convergentdivergent of wherein said ngram sign speech depends on the described number of the atomic unit among the described ngram.

10. system as claimed in claim 8, the described probability through convergentdivergent of wherein said ngram sign speech is x ^{n}, wherein x is the described probability of described ngram sign speech, and n is the described number of the atomic unit among the described ngram.

11. system as claimed in claim 8, the described probability through convergentdivergent of wherein said ngram sign speech is x ^{1+k (n1)}, wherein x is the described probability of described ngram sign speech, n is the described number of the atomic unit among the described ngram, and k is constant and 0≤
 k≤
1.
 k≤

12. system as claimed in claim 8 further comprises:
Dispenser, described dispenser receive a plurality of symbols and use described probability through convergentdivergent that described a plurality of symbol segmentation are become speech.

13. system as claimed in claim 8 further comprises the one or more computing machines that can operate executable operations, described operation comprises:

Identify the ngram of less level, the ngram of described less level obtains from described ngram; The probability of each the corresponding sign speech among the ngram of reception and described less level; The described probability that the combination of the described probability of described ngram sign speech and the ngram of described less level is identified speech compares;
AndWhen the described probability of the probability of the combination sign speech of the ngram of less level and described ngram sign speech differs the assign thresholds amount, revise the corresponding described convergentdivergent weight of described probability that identifies speech with described ngram.


14. system as claimed in claim 8 further comprises the one or more computing machines that can operate executable operations, described operation comprises:

Receive the probability through convergentdivergent of ngram sign speech; Determine the probability through convergentdivergent of the ngram sign speech of less level, the ngram of described less level obtains from described ngram;
AndWhen the described probability through convergentdivergent that identifies speech through the probability of convergentdivergent and described ngram of the combination sign speech of the ngram of less level differs the assign thresholds amount, remove described ngram from dictionary.


15. one kind visibly is stored in computer program on the computerreadable medium, that comprise instruction, described instruction can be operated and impel programmable processor:

Receive the probability of ngram sign speech; Determine the number of the atomic unit among the corresponding ngram; Described number according to the atomic unit among the described ngram identifies the convergentdivergent weight;
AndThe described probability that described convergentdivergent weight is applied to described ngram sign speech determines that described ngram identifies the probability through convergentdivergent of speech.


16. computer program as claimed in claim 15, the described probability through convergentdivergent of wherein said ngram sign speech depends on the described number of the atomic unit among the described ngram.

17. computer program as claimed in claim 15, the described probability through convergentdivergent of wherein said ngram sign speech is x ^{n}, wherein x is the described probability of described ngram sign speech, and n is the described number of the atomic unit among the described ngram.

18. computer program as claimed in claim 15, the described probability through convergentdivergent of wherein said ngram sign speech is x ^{1+k (n1)}, wherein x is the described probability of described ngram sign speech, n is the described number of the atomic unit among the described ngram, and k is constant and 0≤
 k≤
1.
 k≤

19. computer program as claimed in claim 15 further comprises instruction, described instruction can be operated and impel programmable processor:

Receive a plurality of symbols;
AndUse described probability that described a plurality of symbol segmentation are become speech through convergentdivergent.


20. computer program as claimed in claim 15 further comprises instruction, described instruction can be operated and impel programmable processor:

Identify the ngram of less level, the ngram of described less level obtains from described ngram; The probability of each the corresponding sign speech among the ngram of reception and described less level; The described probability that the combination of the described probability of described ngram sign speech and the ngram of described less level is identified speech compares;
AndWhen the described probability of the probability of the combination sign speech of the ngram of less level and described ngram sign speech differs the assign thresholds amount, revise the corresponding described convergentdivergent weight of described probability that identifies speech with described ngram.


21. computer program as claimed in claim 15 further comprises instruction, described instruction can be operated and impel programmable processor:

Receive the probability through convergentdivergent of ngram sign speech; Determine the probability through convergentdivergent of the ngram sign speech of less level, the ngram of described less level obtains from described ngram;
AndWhen the described probability through convergentdivergent that identifies speech through the probability of convergentdivergent and described ngram of the combination sign speech of the ngram of less level differs the assign thresholds amount, remove described ngram from dictionary.


22. a system comprises:

Be used to receive the device that ngram identifies the probability of speech; Be used for determining the device of number of the atomic unit of corresponding ngram; Be used for identifying the device of convergentdivergent weight according to the described number of the atomic unit of described ngram;
AndBe used for the described probability that described convergentdivergent weight is applied to described ngram sign speech is determined that described ngram identifies the device through the probability of convergentdivergent of speech.


23. the system as claimed in claim 22, the described probability through convergentdivergent of wherein said ngram sign speech depends on the described number of the atomic unit among the described ngram.

24. the system as claimed in claim 22, the described probability through convergentdivergent of wherein said ngram sign speech is x ^{n}, wherein x is the described probability of described ngram sign speech, and n is the described number of the atomic unit among the described ngram.

25. the system as claimed in claim 22, the described probability through convergentdivergent of wherein said ngram sign speech is x ^{1+k (n1)}, wherein x is the described probability of described ngram sign speech, n is the described number of the atomic unit among the described ngram, and k is constant and 0≤
 k≤
1.
 k≤
Specification(s)