Correction menu enrichment with alternate choices and generation of choice lists in multi-pass recognition systems

• US 9,858,038 B2
• Filed: 02/01/2013
• Issued: 01/02/2018
• Est. Priority Date: 02/01/2013
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Nuance Communications, Inc. (Microsoft Corporation)
• Inventors: Divay, Olivier, Dubach, Joev, Nagesha, Venkatesh, Gold, Allan
• Primary Examiner(s): Yi, Rinna
• Application Number:
  US13/756,862
• Publication Number:
  US 20140223310A1
• Time in Patent Office: 1,796 Days
• Field of Search: 715706, 715709, 715727, 715728, 715762, 715763, 715771, 715861, 704231, 704233, 704235, 704246, 704260, 704270, 7042701, 704275, 704E11011
• CPC Class Codes: G06F 3/167 Audio in a user interface, e.g. using voice commands for navigating, audio feedback
• CPC Class Codes: G06F 40/232 Orthographic correction, e.g. spell checking or vowelisation
• CPC Class Codes: G10L 15/065 Adaptation
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue
• CPC Class Codes: G10L 2015/221 Announcement of recognition results

Abstract

A method is described for user correction of speech recognition results. A speech recognition result for a given unknown speech input is displayed to a user. A user selection is received of a portion of the recognition result needing to be corrected. For each of multiple different recognition data sources, a ranked list of alternate recognition choices is determined which correspond to the selected portion. The alternate recognition choices are concatenated or interleaved together and duplicate choices removed to form a single ranked output list of alternate recognition choices, which is displayed to the user. The method may be adaptive over time to derive preferences that can then be leveraged in the ordering of one choice list or across choice lists.

Claims

1. A computer-implemented method employing at least one hardware implemented computer processor for user correction of speech recognition results comprising:

displaying, via a user interface, to a user a speech recognition result for a given unknown speech input;

receiving, via the user interface, a user selection of a portion of the speech recognition result needing to be corrected;

for each of a plurality of different recognition post-processing tools, determining a respective ranked list of alternate recognition choices corresponding to the user selection of the portion of the speech recognition result needing to be corrected, the plurality of different recognition post-processing tools including at least two of text formatting, homophone data, and word compounding;

combining the alternate recognition choices determined by the each of the plurality of different recognition post-processing tools together to form a single ranked output list of the alternate recognition choices;

displaying, via the user interface, the single ranked output list to the user for substituting the user selection of the portion of the speech recognition result needing to be corrected with one of the alternate recognition choices, wherein determining the respective ranked list for each of the plurality of different recognition post-processing tools occurs during a limited time period; and

when each of the respective ranked lists fails to include a minimum number of alternate recognition choices and time allows, using some or all of the different recognition post-processing tools to determine additional ranked lists of alternate recognition choices for a next-best recognition choice corresponding to the user selection of the portion of the speech recognition result needing to be corrected; and

further combining the additional alternate recognition choices together into the output list of alternate recognition choices.

2. The method according to claim 1, further comprising:

receiving a user selection of one of the alternate recognition choices in the output list; and

substituting the selected choice for the selected portion in the speech recognition result.

3. The method according to claim 1, wherein the method is adaptive over time to be speaker dependent.

4. The method according to claim 3, wherein the adaptation is based on one or more of past user corrections, user-specific document data, and a user-specific usage corpora.

5. The method according to claim 1, wherein combining the alternate recognition choices includes interleaving and concatenating the alternate recognition choices and removing duplicate choices.

6. The method according to claim 5, wherein interleaving and concatenating the alternate recognition choices places a first choice from each ranked list after a first choice from the output list, and then ranks additional alternate recognition choices based on probability relative to the first choice in their original ranked list.

7. The method according to claim 5, wherein interleaving and concatenating the alternate recognition choices is based on round robin alternation through the ranked lists in a desired order.

8. The method according to claim 5, wherein interleaving and concatenating the alternate recognition choices is directly based on the associated probabilities of the alternate recognition choices.

9. The method according to claim 5, wherein interleaving and concatenating the alternate recognition choices is directly based on the associated probabilities of the alternate recognition choices, except that a desired top alternate recognition choice is boosted to the top of the output list.

10. The method according to claim 1, wherein the different post-processing tools include speech recognition results from sequential speech recognition passes in a multi-pass speech recognition arrangement.

11. The method according to claim 1, wherein the different post-processing tools include speech recognition results from parallel speech recognition passes in a multi-pass speech recognition arrangement.

12. The method according to claim 1, wherein a given post-processing tool may be disabled for a given user selection.

13. The method according to claim 1, wherein an ordering rule may be used for combining the alternate recognition choices.

14. The method according to claim 13, wherein the ordering rule may reflect an ordering of the different post-processing tools.

15. The method according to claim 1, wherein the alternate recognition choices are constrained by one or more language models.

16. The method according to claim 1, wherein enablement and ordering of the different recognition post-processing tools depends on one or more of language, country, topic, and application.