Generation of speech recognition grammars for conducting searches

• US 7,840,405 B1
• Filed: 03/13/2008
• Issued: 11/23/2010
• Est. Priority Date: 03/18/2003
• Status: Active Grant

• Current Assignee: A9.com Incorporated (Amazon.com, Inc.)
• Original Assignee: A9.com Incorporated (Amazon.com, Inc.)
• Inventors: Lee, Nicholas J., Schoenbaum, Ronald J., Frederick, Robert
• Primary Examiner(s): Hudspeth; David R
• Assistant Examiner(s): Neway; Samuel G
• Application Number:
  US12/048,091
• Time in Patent Office: 985 Days
• Field of Search: None
• US Class Current: 704/243
• CPC Class Codes: G06F 16/3329 Natural language query formulation or dialogue systems
• CPC Class Codes: G10L 15/19 Grammatical context, e.g. disambiguation of the recognition hypotheses based on word sequence rules
• CPC Class Codes: G10L 15/26 Speech to text systems G10L15/08 takes precedence

Abstract

Various processes are disclosed for conducting database searches by voice. One such process enables a user to efficiently submit a search query by partially spelling the search query (either on a telephone keypad or via voice utterances) and uttering the full search query. Also disclosed are various processes for generating speech recognition grammars for interpreting utterances of search queries. In one such process, search queries are selected from a search query log for incorporation into a speech recognition grammar. The search query log may include or consist of search queries specified by users without the use of voice.

Claims

1. A method of generating speech recognition grammars for conducting voice-based database searches, the method comprising:

maintaining a query log that stores search queries submitted to a query server by users;

selecting a plurality of search queries from the query log for incorporation into a speech recognition grammar, wherein selecting the plurality of search queries from the query log comprises taking into consideration frequencies with which specific search queries were submitted to the query server by users via a web browser without the use of voice;

incorporating the plurality of search queries into at least one speech recognition grammar that specifies valid utterances for conducting searches; and

using the at least one speech recognition grammar to interpret utterances of search queries by users;

whereby information about search queries submitted by users without the use of voice is used to select, for inclusion in the speech recognition grammar, search queries likely to be used for searching by voice.

2. The method of claim 1, wherein selecting the plurality of search queries from the query log comprises taking into consideration post-search actions performed by users who submitted specific search queries.

3. The method of claim 1, wherein selecting the plurality of search queries from the query log comprises taking into consideration frequencies with which specific search queries were submitted within a selected time period.

4. The method of claim 1, wherein selecting the plurality of search queries from the query log comprises ignoring search query submissions that produced a null query result.

5. The method of claim 1, wherein the plurality of search queries include search phrases submitted by users.

6. The method of claim 1, wherein incorporating the plurality of search queries into at least one speech recognition grammar comprises applying transformations to specific terms in the search queries.

7. The method of claim 1, wherein incorporating the plurality of search queries into at least one speech recognition grammar comprises generating a master grammar set, and subdividing the master grammar set based on the first N characters of search strings contained therein to generate multiple speech recognition grammars.

8. The method of claim 7, wherein the master grammar set is subdivided such that each speech recognition grammar corresponds to a different respective sequence of N keys of a telephone keypad.

9. The method of claim 7, wherein the master grammar set is subdivided so as to provide fault tolerance against misidentification of an utterance of a first character as being an utterance of a second character that does not appear on a common telephone key as the first character.

10. The method of claim 1, further comprising updating the at least one speech recognition grammar periodically based on an analysis of a most recent set of data contained in the query log.

11. A system that operates according to the method of claim 1.

12. A system that supports voice-based database searches, the system comprising:

a query server system that processes, and maintains a log of, search queries submitted by users over a computer network to search a database;

a grammar generation component that selects search queries from the query log to include within speech recognition grammars based, at least in part, on data regarding how frequently particular search queries are submitted to the query server system without the use of voice; and

a speech recognition system that uses the speech recognition grammars generated by the grammar generation component to interpret utterances of search queries by users of the query server system;

whereby the system uses information about search queries submitted without the use of voice to select, for inclusion in the speech recognition grammars, search queries deemed likely to be used to conduct voice-based searches.

13. The system of claim 12, wherein the query server system processes both voice based and non-voice-based search query submissions, and the grammar generation component generates the speech recognition grammars based primarily or entirely on non-voice-based search query submissions contained in the query log.

14. The system of claim 12, wherein the grammar generation component selects search queries from the query log to include within speech recognition grammars based further on post-search actions performed by users.

15. The system of claim 12, wherein the grammar generation component applies transformations to the selected search queries to generate specific entries to include in the speech recognition grammars.

16. The system of claim 12, wherein the grammar generation component substantially ignores search query submissions that produced a null query result.

17. The system of claim 12, wherein the search queries selected by grammar generation component to include within the speech recognition grammars include search phrases submitted by users.

18. The system of claim 12, wherein the grammar generation component generates a master grammar set, and subdivides the master grammar set into multiple speech recognition grammars based on initial sets of characters of each entry in the master grammar set.

19. The system of claim 12, wherein the grammar generation component periodically updates a set of speech recognition grammars to reflect newly acquired query log data.