Voice interface for a search engine

• US 8,515,752 B1
• Filed: 03/12/2008
• Issued: 08/20/2013
• Est. Priority Date: 02/07/2001
• Status: Active Grant

• Current Assignee: Google LLC (Alphabet Inc.)
• Original Assignee: Google Inc. (Alphabet Inc.)
• Inventors: Franz, Alexander Mark, Henzinger, Monika H., Brin, Sergey, Milch, Brian Christopher
• Primary Examiner(s): Lerner, Martin
• Application Number:
  US12/047,108
• Time in Patent Office: 1,987 Days
• Field of Search: 704/239, 704/240, 704/252, 704/275, 704/236, 704/270, 707/3, 707/4, 707/5, 707/713, 707/723, 707/748, 707/728, 707/752, 707/765
• US Class Current: 704/236
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue
• CPC Class Codes: G10L 2015/085 Methods for reducing search complexity, pruning
• CPC Class Codes: Y10S 707/99933 Query processing, i.e. searching
• CPC Class Codes: Y10S 707/99935 Query augmenting and refining, e.g. inexact access

Abstract

A system provides search results from a voice search query. The system receives a voice search query from a user, derives one or more recognition hypotheses, each being associated with a weight, from the voice search query, and constructs a weighted boolean query using the recognition hypotheses. The system then provides the weighted boolean query to a search system and provides the results of the search system to a user.

Claims

1. A method comprising:

receiving, by one or more processors in one or more computer devices, a voice search query;

determining, by the one or more processors and from the voice search query, recognition hypotheses for the voice search query,

the recognition hypotheses being associated with statistical weights,

each of the statistical weights reflecting a respective likelihood that a corresponding recognition hypothesis, of the recognition hypotheses, is accurate for the voice search query;

selecting, by the one or more processors and based on the statistical weights, a plurality of the recognition hypotheses as selected recognition hypotheses,

where the selected recognition hypotheses are associated with, respectively, corresponding-statistical weights of the statistical weights, and

where a quantity of the selected recognition hypotheses is determined based on a quantity of terms included in one of the selected recognition hypotheses and a maximum quantity of terms to be included in a weighted boolean query;

constructing, by the one or more processors, the weighted boolean query,

the weighted boolean query including the selected recognition hypotheses weighted by the corresponding statistical weights;

obtaining, by the one or more processors, results associated with the weighted boolean query; and

providing, by the one or more processors, the results for display.

2. The method of claim 1, where determining the recognition hypotheses includes:

using at least one of a language model, a phonetic dictionary, or an acoustic model to determine the recognition hypotheses from the voice search query.

3. The method of claim 2, further comprising:

updating, based on the voice search query and the results, the at least one of the language model, the phonetic dictionary, or the acoustic models.

4. The method of claim 1, further comprising:

determining a language associated with the voice search query; and

identifying a language model, of a plurality of language models, based on the determined language, and

where determining the recognition hypotheses includes;

determining the recognition hypotheses using the identified language model.

5. The method of claim 2, where determining the recognition hypotheses includes:

determining the statistical weights based on the at least one of the language model, the phonetic dictionary, or the acoustic model used to determine the recognition hypotheses.

6. The method of claim 1, further comprising:

adjusting a ranking of the results based on the corresponding statistical weights.

7. The method of claim 1, further comprising:

grouping sets of the results based on the corresponding statistical weights.

8. The method of claim 1, where providing the results for display further includes:

determining quantities of results associated with respective sets of the results related to the selected recognition hypotheses; and

providing, for display, a set of the results for a recognition hypothesis, of the selected recognition hypotheses, that is associated with one of the quantities of results that satisfies a threshold.

9. The method of claim 1, further comprising:

determining quantities of results associated with respective sets of the results related to the selected recognition hypotheses; and

adjusting the corresponding statistical weights based on the determined quantities of results,

where the method further comprises;

refining the weighted boolean query based on the adjusted corresponding statistical weights, and

obtaining additional results associated with the refined weighted boolean query.

10. A device comprising:

a memory to store instructions; and

at least one processor to execute the instructions to;

receive a voice search query,

determine recognition hypotheses from the voice search query,

determine, for the recognition hypotheses, statistical weights,

each of the statistical weights reflecting a respective likelihood that a corresponding recognition hypothesis, of the recognition hypotheses, is accurate for the voice search query,

select, based on the statistical weights, a plurality of the recognition hypotheses as selected recognition hypotheses,

where the selected recognition hypotheses are associated, respectively, with corresponding statistical weights of the statistical weights, and

where a quantity of the selected recognition hypotheses is based on a quantity of terms included in one of the selected recognition hypotheses and a maximum quantity of terms to be included in a weighted boolean query,

construct the weighted boolean query using the selected recognition hypotheses,

the weighted boolean query including the selected recognition hypotheses weighted by the corresponding statistical weights, and obtain results associated with the weighted boolean query.

11. The device of claim 10, where the at least one processor is further to execute the instructions to:

identify, based on a language associated with the voice search query, a language model from a plurality of language models, and

where the at least one processor, when determining the recognition hypotheses, is further to execute the instructions to;

derive the recognition hypotheses using the identified language model.

12. The device of claim 10, where the at least one processor, when obtaining the results, is further to execute the instructions to at least one of:

adjust a ranking of the results based on the corresponding statistical weights, or

group sets of the results based on the corresponding statistical weights.

13. The device of claim 10, where the at least one processor is further to execute the instructions to:

determine quantities of results associated with respective sets of the results related to the selected recognition hypotheses, and

discard a set of the results related to a recognition hypothesis, of the selected recognition hypotheses, associated with one of the quantities of results that does not satisfy a threshold.

14. The device of claim 10, where the at least one processor is further to execute the instructions to:

determine quantities of results associated with respective sets of the results related to the selected recognition hypotheses, and

adjust the corresponding statistical weights based on the quantities, and

where the at least one processor is further to execute the instructions to;

refine the weighted boolean query based on the adjusted corresponding statistical weights, and

obtain refined results associated with the refined weighted boolean query.

15. A non-transitory computer-readable medium to store instructions executable by one or more processors of one or more devices, the instructions comprising:

one or more instructions to receive a voice search query;

one or more instructions to determine recognition hypotheses based on the voice search query,

where the recognition hypotheses are associated with a plurality of statistical weights that reflect respective likelihoods that the recognition hypotheses are accurate with respect to the voice search query;

one or more instructions to select, based on the plurality of statistical weights and as selected recognition hypotheses, a plurality of the recognition hypotheses,

where a quantity of the selected recognition hypotheses is based on a quantity of terms included in one of the selected recognition hypotheses and a maximum quantity of terms to be included in a weighted boolean query;

one or more instructions to construct the weighted boolean query,

where the weighted boolean query includes the selected recognition hypotheses, weighted by corresponding statistical weights of the plurality of statistical weights; and

one or more instructions to obtain results associated with the weighted boolean query.

16. The non-transitory computer-readable medium of claim 15, where the one or more instructions to determine the recognition hypotheses include:

one or more instructions to use at least one of a language model, a phonetic dictionary, or an acoustic model to determine the recognition hypotheses from the voice search query, and

where the instructions further comprise;

one or more instructions to update, based on the voice search query and the results, the at least one of the language model, the phonetic dictionary, or the acoustic model.

17. The non-transitory computer-readable medium of claim 15, where the instructions further comprise:

one or more instructions to identify a language associated with the voice search query; and

one or more instructions to select a language model, from a plurality of language models, based on the identified language, and

where the one or more instructions to determine the recognition hypotheses include;

one or more instructions to determine the recognition hypotheses using the selected language model.

18. The non-transitory computer-readable medium of claim 15, where the instructions further comprise:

one or more instructions to determine the plurality of statistical weights based on at least one of a language model, a phonetic dictionary, or an acoustic model.

19. The non-transitory computer-readable medium of claim 15, where the one or more instructions to obtain the results include:

one or more instructions to adjust a ranking of the results based on the corresponding statistical weights.

20. The non-transitory computer-readable medium of claim 15, where the one or more instructions to obtain the results include:

one or more instructions to group sets of the results based on the corresponding statistical weights.