Disambiguation of a spoken query term

• US 8,521,526 B1
• Filed: 07/28/2010
• Issued: 08/27/2013
• Est. Priority Date: 07/28/2010
• Status: Expired due to Fees

• Current Assignee: Google LLC (Alphabet Inc.)
• Original Assignee: Google Inc. (Alphabet Inc.)
• Inventors: Lloyd, Matthew I., Schalkwyk, Johan, Risbood, Pankaj
• Primary Examiner(s): Saint Cyr, Leonard
• Application Number:
  US12/845,034
• Time in Patent Office: 1,126 Days
• Field of Search: 704/236, 704/246, 704/247, 704/251, 704/252
• US Class Current: 704/236
• CPC Class Codes: G06F 16/3344 using natural language analysis
• CPC Class Codes: G06F 16/90332 Natural language query formulation or dialogue systems
• CPC Class Codes: G06F 16/9535 Search customisation based on user profiles and personalisation
• CPC Class Codes: G10L 15/01 Assessment or evaluation of speech recognition systems
• CPC Class Codes: G10L 15/183 using context dependencies, e.g. language models
• CPC Class Codes: G10L 15/197 Probabilistic grammars, e.g. word n-grams
• CPC Class Codes: G10L 15/26 Speech to text systems G10L15/08 takes precedence

Abstract

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for processing spoken query terms. In one aspect, a method includes performing speech recognition on an audio signal to select two or more textual, candidate transcriptions that match a spoken query term, and to establish a speech recognition confidence value for each candidate transcription, obtaining a search history for a user who spoke the spoken query term, where the search history references one or more past search queries that have been submitted by the user, generating one or more n-grams from each candidate transcription, where each n-gram is a subsequence of n phonemes, syllables, letters, characters, words or terms from a respective candidate transcription, and determining, for each n-gram, a frequency with which the n-gram occurs in the past search queries, and a weighting value that is based on the respective frequency.

Claims

1. A system comprising:

one or more computers; and

a computer-readable medium coupled to the one or more computers having instructions stored thereon which, when executed by the one or more computers, cause the one or more computers to perform operations comprising;

receiving an audio signal that corresponds to a spoken query term;

performing speech recognition on the audio signal to select two or more textual, candidate transcriptions that match the spoken query term, and to establish a speech recognition confidence value for each candidate transcription;

obtaining a search history for a user who spoke the spoken query term, wherein the search history references one or more past search queries that have been submitted by the user;

generating one or more n-grams from each candidate transcription, wherein each n-gram is a subsequence of n phonemes, syllables, letters, characters, words or terms from a respective candidate transcription;

determining, for each n-gram, a frequency with which the n-gram occurs in the past search queries, and a weighting value that is based on the respective frequency;

generating, for each of the candidate transcriptions, a combined value based on combining the speech recognition confidence value for the candidate transcription with the weighting value for one or more of the n-grams that are generated from the candidate transcription;

selecting an intended query term from among the candidate transcriptions based on the combined values; and

causing a search engine to perform a search query that includes the intended query term.

2. The system of claim 1, wherein the operations further comprise:

determining a context associated with the spoken query term; and

selecting a subset of the past search queries that include contexts which are similar to the context of the spoken query term,

wherein determining, for each n-gram, a frequency with which the n-gram occurs in the past search queries further comprises determining, for each n-gram, a frequency with which the n-gram occurs in the past search queries that are members of the subset only.

3. The system of claim 2, wherein:

determining a context associated with the spoken query term further comprises determining a date that the spoken query term was spoken by the user; and

selecting a subset of the past search queries that include contexts which are similar to the context of the spoken query term further comprises selecting a subset of the past search queries that were submitted by the user on the same date, that were submitted within a predetermined time period before the date, that were submitted on a same day of the week as the date, that were submitted on the same day, week, month or year as the date, that were submitted on a weekend when the date occurs on a weekend, or that were submitted on a weekday when the date occurs on a weekday.

4. The system of claim 2, wherein:

determining a context associated with the spoken query term further comprises determining a time that the spoken query term was spoken by the user; and

selecting a subset of the past search queries that include contexts which are similar to the context of the spoken query term further comprises selecting a subset of the past search queries that were submitted by the user within a predetermined amount of period before the time, that were submitted on a same time of day as the time;

that were submitted on a same minute or hour as the time, that were submitted on during daytime when the time occurs during daytime, or that were submitted during nighttime when the time occurs during nighttime.

5. The system of claim 2, wherein:

determining a context associated with the spoken query term further comprises determining that a device used by the user to enter the spoken query term was docked or holstered when the user spoke the spoken query terms; and

selecting a subset of the past search queries that include contexts which are similar to the context of the spoken query term further comprises selecting a subset of the past search queries that were submitted by the user when a device used by the user to enter the past search queries was docked or holstered when the user entered the past search queries.

6. The system of claim 2, wherein:

determining a context associated with the spoken query term further comprise determining a location of the user when the spoken query term was spoken by the user; and

selecting a subset of the past search queries that include contexts which are similar to the context of the spoken query term further comprises selecting a subset of the past search queries that were submitted by the user when the user was within a predetermined distance from the location, or when the user was in a same geographic region as the location.

7. The system of claim 2, wherein:

determining a context associated with the spoken query term further comprises determining information that specifies a device or a type of device that the user used to enter the spoken query term; and

selecting a subset of the past search queries that include contexts which are similar to the context of the spoken query term further comprises selecting a subset of the past search queries that were submitted by the user using the same device or the same type of device that the user used to enter the spoken query term.

8. The system of claim 2, wherein:

the operations further comprise selecting a subset of the past search queries that were submitted by the user as voice search queries and that generated one or more search results that were selected by the user; and

determining, for each n-gram, a frequency with which the n-gram occurs in the past search queries further comprises determining, for each n-gram, a frequency with which the n-gram occurs in the past search queries that are members of the subset only.

9. The system of claim 2, wherein:

the operations further comprise selecting a subset of the past search queries that were selected by the user from a list of m-best search queries; and

determining, for each n-gram, a frequency with which the n-gram occurs in the past search queries further comprises determining, for each n-gram, a frequency with which the n-gram occurs in the past search queries that are members of the subset only.

10. The system of claim 1, wherein:

the operations further comprise selecting a subset of the past search queries that include query terms that were entered by the user using a keyboard; and

determining, for each n-gram, a frequency with which the n-gram occurs in the past search queries further comprises determining, for each n-gram, a frequency with which the n-gram occurs in the past search queries that are members of the subset only.

11. The system of claim 1, wherein:

the operations further comprise determining one or more categories that are associated with each of the past search queries; and

for each n-gram, determining a frequency with which the n-gram occurs in the past search queries further comprises;

determining a particular category that is associated with the n-gram; and

determining a quantity of the past search queries that are associated the particular category.

12. The system of claim 1, wherein:

the operations further comprise selecting the candidate transcriptions that have an m highest combined values and providing information to the user that references the selected candidate transcriptions that have the m highest combined values; and

selecting the intended query term from among the candidate transcriptions based on the combined values further comprises receiving an indication of one of the selected candidate transcriptions that have the m highest combined values that has been selected by the user.

13. The system of claim 1, wherein selecting an intended query term from among the candidate transcriptions based on the combined values further comprises automatically selecting, as the intended query term, the candidate transcription that has a highest combined value from among the candidate transcriptions.

14. A non-transitory computer storage medium encoded with a computer program, the program comprising instructions that when executed by data processing apparatus cause the data processing apparatus to perform operations comprising:

receiving an audio signal that corresponds to a spoken query term;

performing speech recognition on the audio signal to select two or more textual, candidate transcriptions that match the spoken query term, and to establish a speech recognition confidence value for each candidate transcription;

obtaining a search history for a user who spoke the spoken query term, wherein the search history references one or more past search queries that have been submitted by the user;

generating one or more n-grams from each candidate transcription, wherein each n-gram is a subsequence of n phonemes, syllables, letters, characters, words or terms from a respective candidate transcription;

determining, for each n-gram, a frequency with which the n-gram occurs in the past search queries, and a weighting value that is based on the respective frequency;

generating, for each of the candidate transcriptions, a combined value based on combining the speech recognition confidence value for the candidate transcription with the weighting value for one or more of the n-grams that are generated from the candidate transcription;

selecting an intended query term from among the candidate transcriptions based on the combined values; and

causing a search engine to perform a search query that includes the intended query term.

15. The non-transitory computer storage medium of claim 14, wherein the operations further comprise:

determining a context associated with the spoken query term; and

selecting a subset of the past search queries that include contexts which are similar to the context of the spoken query term,

wherein determining, for each n-gram, a frequency with which the n-gram occurs in the past search queries further comprises determining, for each n-gram, a frequency with which the n-gram occurs in the past search queries that are members of the subset only.

16. The non-transitory computer storage medium of claim 15, wherein:

determining a context associated with the spoken query term further comprises determining a date that the spoken query term was spoken by the user; and

selecting a subset of the past search queries that include contexts which are similar to the context of the spoken query term further comprises selecting a subset of the past search queries that were submitted by the user on the same date, that were submitted within a predetermined time period before the date, that were submitted on a same day of the week as the date, that were submitted on the same day, week, month or year as the date, that were submitted on a weekend when the date occurs on a weekend, or that were submitted on a weekday when the date occurs on a weekday.

17. The non-transitory computer storage medium of claim 15, wherein:

determining a context associated with the spoken query term further comprises determining a time that the spoken query term was spoken by the user; and

selecting a subset of the past search queries that include contexts which are similar to the context of the spoken query term further comprises selecting a subset of the past search queries that were submitted by the user within a predetermined amount of period before the time, that were submitted on a same time of day as the time;

that were submitted on a same minute or hour as the time, that were submitted on during daytime when the time occurs during daytime, or that were submitted during nighttime when the time occurs during nighttime.

18. The non-transitory computer storage medium of claim 15, wherein:

determining a context associated with the spoken query term further comprises determining that a device used by the user to enter the spoken query term was docked or holstered when the user spoke the spoken query terms; and

selecting a subset of the past search queries that include contexts which are similar to the context of the spoken query term further comprises selecting a subset of the past search queries that were submitted by the user when a device used by the user to enter the past search queries was docked or holstered when the user entered the past search queries.

19. The non-transitory computer storage medium of claim 15, wherein:

determining a context associated with the spoken query term further comprise determining a location of the user when the spoken query term was spoken by the user; and

selecting a subset of the past search queries that include contexts which are similar to the context of the spoken query term further comprises selecting a subset of the past search queries that were submitted by the user when the user was within a predetermined distance from the location, or when the user was in a same geographic region as the location.

20. A method comprising:

receiving an audio signal that corresponds to a spoken query term;

performing speech recognition on the audio signal to select two or more textual, candidate transcriptions that match the spoken query term, and to establish a speech recognition confidence value for each candidate transcription;

obtaining a search history for a user who spoke the spoken query term, wherein the search history references one or more past search queries that have been submitted by the user;

generating one or more n-grams from each candidate transcription, wherein each n-gram is a subsequence of n phonemes, syllables, letters, characters, words or terms from a respective candidate transcription;

determining, for each n-gram, a frequency with which the n-gram occurs in the past search queries, and a weighting value that is based on the respective frequency;

generating, by one or more computers, and for each of the candidate transcriptions, a combined value based on combining the speech recognition confidence value for the candidate transcription with the weighting value for one or more of the n-grams that are generated from the candidate transcription;

selecting an intended query term from among the candidate transcriptions based on the combined values; and

causing a search engine to perform a search query that includes the intended query term.