Dynamic language model

• US 9,043,205 B2
• Filed: 03/13/2013
• Issued: 05/26/2015
• Est. Priority Date: 06/21/2012
• Status: Active Grant

• Current Assignee: Google LLC (Alphabet Inc.)
• Original Assignee: Google Inc. (Alphabet Inc.)
• Inventors: Mengibar, Pedro J. Moreno, Cohen, Michael H.
• Primary Examiner(s): Baker, Charlotte M
• Application Number:
  US13/802,414
• Publication Number:
  US 20130346077A1
• Time in Patent Office: 804 Days
• Field of Search: 704/235, 704/231
• US Class Current: 704/235
• CPC Class Codes: G06F 16/211 Schema design and management
• CPC Class Codes: G06F 16/29 Geographical information databases
• CPC Class Codes: G06F 16/3329 Natural language query formulation or dialogue systems
• CPC Class Codes: G06F 16/3344 using natural language analysis
• CPC Class Codes: G06F 16/3346 using probabilistic model
• CPC Class Codes: G06F 16/338 Presentation of query results
• CPC Class Codes: G06F 16/387 using geographical or spatial information, e.g. location
• CPC Class Codes: G10L 15/005 Language recognition
• CPC Class Codes: G10L 15/14 using statistical models, e.g. Hidden Markov Models [HMMs] G10L15/18 takes precedence
• CPC Class Codes: G10L 15/197 Probabilistic grammars, e.g. word n-grams
• CPC Class Codes: G10L 15/24 Speech recognition using non-acoustical features
• CPC Class Codes: G10L 15/26 Speech to text systems G10L15/08 takes precedence
• CPC Class Codes: G10L 2015/0633 using lexical or orthographic knowledge sources
• CPC Class Codes: G10L 2015/081 Search algorithms, e.g. Baum-Welch or Viterbi
• CPC Class Codes: G10L 2015/228 of application context

Abstract

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for speech recognition. One of the methods includes receiving a base language model for speech recognition including a first word sequence having a base probability value; receiving a voice search query associated with a query context; determining that a customized language model is to be used when the query context satisfies one or more criteria associated with the customized language model; obtaining the customized language model, the customized language model including the first word sequence having an adjusted probability value being the base probability value adjusted according to the query context; and converting the voice search query to a text search query based on one or more probabilities, each of the probabilities corresponding to a word sequence in a group of one or more word sequences, the group including the first word sequence having the adjusted probability value.

Claims

1. A method comprising:

receiving a base language model for speech recognition, the base language model including a first word sequence having a base probability value;

receiving a voice search query, the voice search query being associated with a query context;

determining that a customized language model is to be used when the query context satisfies one or more criteria associated with the customized language model;

obtaining the customized language model, the customized language model including the first word sequence having an adjusted probability value, the adjusted probability value being the base probability value adjusted according to the query context; and

converting the voice search query to a text search query based on one or more probabilities, each of the one or more probabilities corresponding to a word sequence in a group of one or more word sequences, the group of one or more word sequences including the first word sequence having the adjusted probability value.

2. The method of claim 1, wherein obtaining the customized language model comprises:

selecting the customized language model from one or more customized language models based on the query context.

3. The method of claim 1, wherein obtaining the customized language model comprises:

adjusting the base language model based on the received query context using one or more language adjustment rules.

4. The method of claim 1, wherein the query context includes a geographic location of a user device from which the voice search query is received.

5. The method of claim 4, comprising:

determining that the word sequence relates to the geographic location, wherein the adjusted probability value includes the base probability value adjusted based on a degree of relationship between the word sequence and the geographic location.

6. The method of claim 5, wherein:

the word sequence includes a name of a feature located at or proximate to the geographic location; and

the degree of relationship is determined based on a distance between the feature and the geographic location in the query context.

7. The method of claim 5, wherein:

the word sequence includes a term of a past query received from the geographic location at a frequency that exceeds a threshold; and

the degree of relationship is determined based on the frequency.

8. The method of claim 4, wherein the geographic location is a geographic area defined by one or more sets of latitude and longitude coordinates.

9. The method of claim 1, wherein the query context includes a user provided identifier for associating with the voice search query.

10. The method of claim 9, comprising:

determining a social group associated with the user provided identifier, the social group including one or more user identifiers that are related to the user provided identifier;

determining a frequency of occurrence of the word sequence using stored query logs that are associated with the user identifiers in the social group; and

adjusting the base probability value based on the frequency of occurrence.

11. The method of claim 10, wherein adjusting the base probability value is further based on a degree of relationship in the social group between each of the user identifiers and the user provided identifier.

12. The method of claim 1, wherein the query context includes a timestamp of the voice search query.

13. The method of claim 12, wherein obtaining the customized language model comprises:

identifying one or more electronic documents, the one or more electronic documents being associated with a number of search queries according to a query log, the number of search queries satisfying a threshold number, the search queries having occurred within a specified time window of the timestamp, each of the one or more electronic documents including the word sequence; and

adjusting the base probability value based on the number of search queries.

14. The method of claim 13, wherein each of the one or more electronic documents is a news article.

15. The method of claim 1, comprising:

providing for display on a display device, as a response to the voice search query, one or more search results of the text search query.

16. The method of claim 1, wherein converting the voice search query to a text search query comprises converting the voice search query to a text search query that includes the first word sequence based on the corresponding adjusted probability value in the customized language model.

17. A system comprising:

one or more computers and one or more storage devices storing instructions that are operable, when executed by the one or more computers, to cause the one or more computers to perform operations comprising;

receiving a base language model for speech recognition, the base language model including a first word sequence having a base probability value;

receiving a voice search query, the voice search query being associated with a query context;

determining that a customized language model is to be used when the query context satisfies one or more criteria associated with the customized language model;

obtaining the customized language model, the customized language model including the first word sequence having an adjusted probability value, the adjusted probability value being the base probability value adjusted according to the query context; and

converting the voice search query to a text search query based on one or more probabilities, each of the one or more probabilities corresponding to a word sequence in a group of one or more word sequences, the group of one or more word sequences including the first word sequence having the adjusted probability value.

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

receiving a base language model for speech recognition, the base language model including a first word sequence having a base probability value;

receiving a voice search query, the voice search query being associated with a query context;

determining that a customized language model is to be used when the query context satisfies one or more criteria associated with the customized language model;

obtaining the customized language model, the customized language model including the first word sequence having an adjusted probability value, the adjusted probability value being the base probability value adjusted according to the query context; and

converting the voice search query to a text search query based on one or more probabilities, each of the one or more probabilities corresponding to a word sequence in a group of one or more word sequences, the group of one or more word sequences including the first word sequence having the adjusted probability value.