Dynamic language model
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.
1. A method comprising:
receiving a voice search query, the voice search query provided by a user to a user device;
determining that a particular customized language model of a plurality of customized language models is to be used to recognize the voice search query, wherein each customized language model includes one or more adjusted probabilities for respective word sequences of the language model, wherein the probabilities are adjusted with respect to corresponding word sequence probabilities in a base language model;
using the particular customized language model to generate a text search query from the voice search query; and
providing for display on the user device, one or more search results responsive to the text search query.
2. The method of claim 1 , wherein each customized language model is generated according one or more language adjustment rules for adjusting the respective probabilities of the base language model based on a particular geographic location.
3. The method of claim 1 , wherein determining that the particular customized language model is to be used comprises determining that the query is received by a user device having a particular geographic location associated with the particular customized language model.
4. The method of claim 3 , wherein adjusting a probability for the particular customized language model includes adjusting a probability of a particular word sequence based on a degree of relationship between the particular word sequence and a geographic location of the user device.
5. The method of claim 4 , wherein:
the particular 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.
6. The method of claim 1 , wherein the determining that the particular customized language model is to be used comprises determining one or more terms of the query and occurrences of the one or more terms by users sharing a social group with the user.
7. The method of claim 6 , wherein adjusting a probability for the particular customized language model includes determining a frequency of occurrence of the word sequence using stored query logs that are associated with users of the social group and adjusting a probability of a particular word sequence based on the frequency.
8. The method of claim 1 , wherein the determining that the particular customized language model is to be used comprises determining an association between the received query and one or more current events.
9. 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 voice search query, the voice search query provided by a user to a user device;
determining that a particular customized language model of a plurality of customized language models is to be used to recognize the voice search query, wherein each customized language model includes one or more adjusted probabilities for respective word sequences of the language model, wherein the probabilities are adjusted with respect to corresponding word sequence probabilities in a base language model;
using the particular customized language model to generate a text search query from the voice search query; and
providing for display on the user device, one or more search results responsive to the text search query.
10. The system of claim 9 , wherein each customized language model is generated according one or more language adjustment rules for adjusting the respective probabilities of the base language model based on a particular geographic location.
11. The system of claim 9 , wherein determining that the particular customized language model is to be used comprises determining that the query is received by a user device having a particular geographic location associated with the particular customized language model.
12. The system of claim 11 , wherein adjusting a probability for the particular customized language model includes adjusting a probability of a particular word sequence based on a degree of relationship between the particular word sequence and a geographic location of the user device.
13. The system of claim 12 , wherein:
the particular 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.
14. The system of claim 9 , wherein the determining that the particular customized language model is to be used comprises determining one or more terms of the query and occurrences of the one or more terms by users sharing a social group with the user.
15. The system of claim 14 , wherein adjusting a probability for the particular customized language model includes determining a frequency of occurrence of the word sequence using stored query logs that are associated with users of the social group and adjusting a probability of a particular word sequence based on the frequency.
16. The system of claim 9 , wherein the determining that the particular customized language model is to be used comprises determining an association between the received query and one or more current events.
17. One or more non-transitory computer storage media 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 voice search query, the voice search query provided by a user to a user device;
determining that a particular customized language model of a plurality of customized language models is to be used to recognize the voice search query, wherein each customized language model includes one or more adjusted probabilities for respective word sequences of the language model, wherein the probabilities are adjusted with respect to corresponding word sequence probabilities in a base language model;
using the particular customized language model to generate a text search query from the voice search query; and
providing for display on the user device, one or more search results responsive to the text search query.
18. The one or more non-transitory computer storage media of claim 17 , wherein each customized language model is generated according one or more language adjustment rules for adjusting the respective probabilities of the base language model based on a particular geographic location.
19. The one or more non-transitory computer storage media of claim 17 , wherein determining that the particular customized language model is to be used comprises determining that the query is received by a user device having a particular geographic location associated with the particular customized language model.
20. The one or more non-transitory computer storage media of claim 19 , wherein adjusting a probability for the particular customized language model includes adjusting a probability of a particular word sequence based on a degree of relationship between the particular word sequence and a geographic location of the user device.
21. The one or more non-transitory computer storage media of claim 20 , wherein:
the particular 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.
22. The one or more non-transitory computer storage media of claim 17 , wherein the determining that the particular customized language model is to be used comprises determining one or more terms of the query and occurrences of the one or more terms by users sharing a social group with the user.
23. The one or more non-transitory computer storage media of claim 22 , wherein adjusting a probability for the particular customized language model includes determining a frequency of occurrence of the word sequence using stored query logs that are associated with users of the social group and adjusting a probability of a particular word sequence based on the frequency.
24. The one or more non-transitory computer storage media of claim 17 , wherein the determining that the particular customized language model is to be used comprises determining an association between the received query and one or more current events.