Generating confidence scores from word lattices

• US 7,562,010 B1
• Filed: 06/23/2006
• Issued: 07/14/2009
• Est. Priority Date: 03/29/2002
• Status: Expired due to Term

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: AT&T Intellectual Property I LP (AT&T, Inc.)
• Inventors: Gretter, Roberto, Riccardi, Giuseppe
• Primary Examiner(s): Hudspeth, David R
• Assistant Examiner(s): Albertalli, Brian L
• Application Number:
  US11/426,225
• Time in Patent Office: 1,117 Days
• Field of Search: None
• US Class Current: 704/9
• CPC Class Codes: G10L 15/08 Speech classification or search

Abstract

Systems and methods for determining word confidence scores. Speech recognition systems generate a word lattice for speech input. Posterior probabilities of the words in the word lattice are determined using a forward-backward algorithm. Next, time slots are defined for the word lattice, and for all transitions that at least partially overlap a particular time slot, the posterior probabilities of transitions that have the same word label are combined for those transitions. The combined posterior probabilities are used as confidence scores. A local entropy can be computed on the competitor transitions of a particular time slot and also used as a confidence score.

Claims

1. A system for determining confidence scores for the labels in a lattice that has one or more transitions that each have a label, the system comprising:

a module configured to determine a posterior probability for each transition of the lattice;

a module configured to define one or more time slots for the lattice, wherein one or more transitions is at least partially included in each one or more time slots; and

a module configured to, for each transition that is at least partially included in a particular time slot, combine the posterior probabilities of all transitions that have the same label to obtain a confidence score for that label.

2. A system as defined in claim 1, wherein the module configured to determine a posterior probability for each transition of the lattice further applies a forward-backward algorithm to the lattice.

3. A system as defined in claim 1, wherein the module configured to define one or more time slots for the lattice further define one or more time slots independently of a topology of the lattice.

4. A system as defined in claim 1, wherein each label is a word or a phoneme.

5. A system as defined in claim 1, wherein the one or more time slots correspond to transitions of a best path in the lattice.

6. A system as defined in claim 1, further comprising a module configured to identify ally transitions in each time slot and identify competitor transitions in each time slot.

7. A system as defined in claim 6, further comprising a module configured to compute a local entropy on competitor transitions in each time slot.

8. A system for determining word confidence scores for word labels in a word lattice generated from speech input and having one or more transitions, the system comprising:

a module configured to apply a forward-backward algorithm to the word lattice to determine a posterior probability for each transition of the word lattice, wherein each transition has a word label;

a module configured to define one or more time slots, wherein each time slot has a starting time frame and an ending time frame; and

a module configured to, for transitions of the word lattice that overlap a particular time slot, combine the posterior probabilities of all transitions that have the same word label, wherein the combined posterior probabilities are a confidence score for the word labels of those transitions.

9. A system as defined in claim 8, wherein the module configured to apply a forward-backward algorithm to the word lattice further combines local likelihoods in order to compute the posterior probabilities of each transition.

10. A system as defined in claim 8, wherein the module configured to define one or more time slots further define one or more time slots using starting time frames and ending time frames that correspond to transitions of a best path of the word lattice.

11. A system as defined in claim 8, wherein the module configured to define one or more time slots further defines one or more time slots independently of a topology of the word lattice.

12. A system as defined in claim 8, wherein the module configured to combine the posterior probabilities of all transitions that have the same word label further identifies both ally transitions and competitor transitions for each transition that overlaps the particular time slot, wherein ally transitions have the same word label.

13. A system as defined in claim 8, wherein the module configured to combine the posterior probabilities of all transitions that have the same word label further sums posterior probabilities for ally transitions.

14. A system as defined in claim 13, further comprising a module configured to compute a local entropy on competitor transitions of the particular time slot.

15. A system for computing word confidence scores for the word labels that are included in a word lattice generated from a speech recognition system that does not require training, the word lattice having one or more transitions that each have a word label, the system comprising:

a module configured to determine one or more time slots for the word lattice, wherein each transition of the word lattice overlaps at least one time slot and wherein each transition in the word lattice has a posterior probability;

a module configured to, for each time slot;

identify transitions that have the same word label; and

sum posterior probabilities of transitions that have the same word label to obtain a word confidence score for those word labels.

17. A system as defined in claim 15, wherein the module configured to identify transitions that have the same word label further identifies transitions that have the same word label from a set of transitions, wherein the set of transitions includes all transitions that at least partially overlap the time slot.

18. A system as defined in claim 15, wherein the module configured to sum posterior probabilities of transitions further sums posterior probabilities of ally transitions.

19. A system as defined in claim 15, further comprising a module configured to compute a local entropy on competitor transitions of a particular time slot, wherein competitor transitions have different word labels.

20. A system as defined in claim 15, wherein the module configured to determine one or more time slots for the word lattice further determines one or more time slots independently of a topology of the word lattice.

16. A system as defined in 15, wherein the module configured to determine one or more time slots for the word lattice further:

applies a forward-backward algorithm to the word lattice; and

combines local likelihoods to determine posterior probabilities for each word label of each transition of the word lattice.

21. A system for determining word confidence scores for word labels in a word lattice generated by a speech recognition system, wherein the word lattice has one or more transitions that each have a word label, the system comprising:

a module configured to define one or more time slots that correspond to transitions of a best path of the word lattice, wherein each time slot has a starting time frame and an ending time frame;

a module configured to identify ally transitions and competitor transitions for each time slot, wherein ally transitions have the same word label and wherein competitor transitions have different labels; and

a module configured to compute a local entropy on competitor transitions in each time slot, wherein the local entropy is a confidence score for the word labels of the competitor transitions.