Query rewrite corrections

• US 10,127,909 B2
• Filed: 01/22/2018
• Issued: 11/13/2018
• Est. Priority Date: 08/29/2014
• Status: Active Grant

• Current Assignee: Google LLC (Alphabet Inc.)
• Original Assignee: Google LLC (Alphabet Inc.)
• Inventors: Skobeltsyn, Gleb, Cherepanov, Evgeny A., Behzadi, Behshad
• Primary Examiner(s): Baker, Charlotte M
• Application Number:
  US15/877,173
• Publication Number:
  US 20180166079A1
• Time in Patent Office: 295 Days
• Field of Search: 704257, 704270, 704231, 704236, 704252, 704251, 704 9, 704E15045, 706 46
• CPC Class Codes: G06F 16/3329 Natural language query formulation or dialogue systems
• CPC Class Codes: G06F 16/3334 Selection or weighting of terms from queries, including natural language queries
• CPC Class Codes: G06F 16/685 using automatically derived transcript of audio data, e.g. lyrics speech recognition G10L15/00
• CPC Class Codes: G06F 40/211 Syntactic parsing, e.g. based on context-free grammar [CFG] or unification grammars
• CPC Class Codes: G06F 40/279 Recognition of textual entities
• CPC Class Codes: G10L 15/01 Assessment or evaluation of speech recognition systems
• CPC Class Codes: G10L 15/08 Speech classification or search
• CPC Class Codes: G10L 15/187 Phonemic context, e.g. pronunciation rules, phonotactical constraints or phoneme n-grams
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue
• CPC Class Codes: G10L 2015/223 Execution procedure of a spoken command

Abstract

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for natural language processing. One of the methods includes receiving a first voice query; generating a first recognition output; receiving a second voice query; determining from a recognition of the second voice query that the second voice query triggers a correction request; using the first recognition output and the second recognition to determine a plurality of candidate corrections; scoring each candidate correction; and generating a corrected recognition output for a particular candidate correction having a score that satisfies a threshold value.

Claims

1. A method comprising:

receiving a first voice input;

generating a first recognition output for the first voice input;

receiving a second voice input;

determining that a correction request for the first recognition output is triggered by the second voice input, wherein the determining comprises;

parsing the second voice input using a plurality of different correction request grammars, and

determining that the correction request for the first recognition output is triggered based on the parsing; and

generating a corrected recognition output in response to the correction request.

2. The method of claim 1, wherein in response to determining that the second voice input triggers a correction request, determining a plurality of candidate corrections.

3. The method of claim 2, wherein one or more candidate corrections are generated using a plurality of recognition hypotheses for the second voice input.

4. The method of claim 3, wherein candidate corrections are generated by substituting portions of the recognition hypotheses into the first recognition output for the first voice input.

5. The method of claim 2, wherein one or more candidate corrections are generated using an obtained misrecognition index.

6. The method of claim 5, wherein the misrecognition index provides a lookup of possible phonetically similar values to particular text given a particular context.

7. The method of claim 5, wherein the misrecognition index is generated from typed queries.

8. The method of claim 5, wherein the misrecognition index maintains computed top-N lists ranked based on phonetic similarity to actual phonetic representations.

9. The method of claim 2, wherein one or more candidate corrections are generated using spelling analysis.

10. The method of claim 9, wherein candidate corrections are generated using spelling analysis when the second voice input includes a recitation of spelling for a misrecognized portion of the first recognition output.

11. The method of claim 2, wherein one or more candidate corrections are generated using a category index.

12. The method of claim 2, wherein each candidate correction is scored based on a confidence measure of a specific applied grammar.

13. The method of claim 2, wherein the voice input is a voice query and wherein each candidate correction is scored based on a measure of query quality indicating a probability that the user intended to issue the candidate correction as a query.

14. The method of claim 2, wherein the voice input is a voice query and wherein each candidate correction is scored based on a measure of replacement query quality indicating how the query would improve if the candidate correction replaces the original misrecognized query.

15. The method of claim 2, wherein each candidate correction is scored based on a combination of a plurality of scoring factors.

16. A system comprising:

a user device; and

one or more computers configured to interact with the user device and to perform operations comprising;

receiving a first voice input;

generating a first recognition output for the first voice input;

receiving a second voice input;

determining that a correction request for the first recognition output is triggered by the second voice input, wherein the determining comprises;

parsing the second voice input using a plurality of different correction request grammars, and

determining that the correction request for the first recognition output is triggered based on the parsing; and

generating a corrected recognition output in response to the correction request.

17. The system of claim 16, wherein in response to determining that the second voice input triggers a correction request, determining a plurality of candidate corrections.

18. The system of claim 17, wherein one or more candidate corrections are generated using a plurality of recognition hypotheses for the second voice input.

19. The system of claim 18, wherein candidate corrections are generated by substituting portions of the recognition hypotheses into the first recognition output for the first voice input.

20. One or more non-transitory computer-readable storage media encoded with instructions that, when executed by one or more computers, cause the one or more computers to perform operations comprising:

receiving a first voice input;

generating a first recognition output for the first voice input;

receiving a second voice input;

determining that a correction request for the first recognition output is triggered by the second voice input, wherein the determining comprises;

parsing the second voice input using a plurality of different correction request grammars, and

determining that the correction request for the first recognition output is triggered based on the parsing; and

generating a corrected recognition output in response to the correction request.