Generation and selection of speech recognition grammars for conducting searches

Abstract

Various processes are disclosed for generating and selecting speech recognition grammars for conducting searches by voice. In one such process, search queries are selected from a search query log for incorporation into speech recognition grammar. The search query log may include or consist of search queries specified by users without the use of voice. Another disclosed process enables a user to efficiently submit a search query by partially spelling the search query (e.g., on a telephone keypad or via voice utterances) and uttering the full search query. The user's partial spelling is used to select a particular speech recognition grammar for interpreting the utterance of the full search query.

1 . A computer-implemented method of supporting voice-based searches, comprising:
 generating a plurality of speech recognition grammars, each of which specifies a respective set of valid search query utterances for conducting voice-based searches, wherein generating the plurality of speech recognition grammars comprises selecting search queries from log of search queries submitted by users;   receiving, from a telephonic device of a user conducting a search, (1) a representation of a partial spelling by the user of a search query, and (2) a representation of an utterance by the user of the search query;   selecting a particular speech recognition grammar from said plurality of speech recognition grammars based on the representation of the partial spelling of the search query; and   interpreting the utterance of the search query with the selected speech recognition grammar;   said method performed by a computerized system that comprises one or more computers.   
     
     
         2 . The method of  claim 1 , wherein generating the plurality of speech recognition grammars comprises generating a master grammar set that comprises a plurality of textual entries, and subdividing the master grammar set into said plurality of speech recognition grammars by grouping together textual entries based on initial characters thereof. 
     
     
         3 . The method of  claim 2 , wherein generating the master grammar set comprises selecting, for inclusion as textual entries in said master grammar set, particular search queries from said log. 
     
     
         4 . The method of  claim 1 , wherein generating the plurality of speech recognition grammars comprises selecting search queries from said log based at least partly on how frequently particular search queries have been submitted over a period of time without use of voice. 
     
     
         5 . The method of  claim 1 , wherein selecting search queries from the log comprises taking into consideration post-search actions performed by users who submitted specific search queries. 
     
     
         6 . The method of  claim 1 , wherein generating the plurality of speech recognition grammars comprises applying transformations to specific terms of search queries selected from the log. 
     
     
         7 . The method of  claim 1 , wherein the representation of the partial spelling of the search query is an indication of a sequence of telephone keys selected by the user to spell a portion of the search query. 
     
     
         8 . The method of  claim 7 , wherein the partial spelling is based on a one-key-depression-per-character entry method in which the user does not uniquely identify any character of the search query. 
     
     
         9 . The method of  claim 1 , wherein the representation of the partial spelling of the search query is a representation of sequence of character utterances of a subset of characters of the search query, and wherein the particular speech recognition grammar is selected based on the sequence of character utterances. 
     
     
         10 . The method of  claim 1 , wherein interpreting the utterance of the search query comprises converting the utterance into a textual representation of the search query using the selected speech recognition grammar, and wherein the method further comprises submitting the textual representation of the search query to a query server for execution. 
     
     
         11 . A system for supporting voice-based searches, comprising:
 a query log that stores search queries submitted by users to a search engine;   a grammar generator that generates speech recognition grammars in part by selecting search queries from the query log, said speech recognition grammars comprising textual entries that represent valid search query utterances for conducting searches by voice; and   an automated speech recognition system that uses the speech recognition grammars to convert search query utterances of users into textual search queries for execution, the automated speech recognition system configured select a speech recognition grammar for interpreting a user's utterance of a search query based on user-supplied representation of a portion of the search query.   
     
     
         12 . The system of  claim 11 , wherein the user-supplied representation of the portion of the search query is a representation of a user-entered sequence of telephone keys corresponding to a subset of characters of the search query. 
     
     
         13 . The system of  claim 12 , wherein the automated speech recognition system is configured to use the sequence of telephone keys to select said speech recognition grammar from a plurality of speech recognition grammars pre-generated by the grammar generator. 
     
     
         14 . The system of  claim 11 , wherein the user-supplied representation of a portion of the search query is a representation of a sequence of character utterances of a subset of characters of the search query 
     
     
         15 . The system of  claim 11 , wherein the log is based at least partly on non-voice-based submissions of search queries of users. 
     
     
         16 . The system of  claim 11 , wherein the grammar generator is configured to generate a master grammar set that comprises a plurality of textual entries, and to subdivide the master grammar set into said plurality of speech recognition grammars by grouping together textual entries based on initial characters thereof, and wherein the automated speech recognition system is configured to select from said plurality of speech recognition grammars based on partial search query spellings supplied by users conducting searches. 
     
     
         17 . The system of  claim 11 , wherein the grammar generator is configured to select search queries from the log for inclusion in the speech recognition grammars based at least partly on how frequently particular search queries were submitted to the search engine without use of voice. 
     
     
         18 . The system of  claim 11 , wherein the grammar generator comprises a general purpose computer that executes software. 
     
     
         19 . Non-transitory computer storage which stores executable code that directs a computer system to at least:
 generate a plurality of speech recognition grammars in part by selecting search queries from a log of search queries submitted by users, each speech recognition grammar specifying a respective set of valid search query utterances for conducting voice-based searches; and   select, from said plurality of speech recognition grammars, a particular speech recognition grammar to use to interpret an utterance of a search query by a user, such that the particular speech recognition grammar is selected based on a representation of a partial spelling of the search query by the user.   
     
     
         20 . The computer storage of  claim 19 , wherein the representation of the partial spelling is a representation of a sequence of telephone keys corresponding to a subset of characters of the search query. 
     
     
         21 . The computer storage of  claim 19 , wherein the representation of the partial spelling is a representation of a sequence character utterances of a subset of characters of the search query.