Intelligent speech recognition of incomplete phrases

• US 7,949,536 B2
• Filed: 08/31/2006
• Issued: 05/24/2011
• Est. Priority Date: 08/31/2006
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Plumpe, Michael D., Schmid, Philipp Heinz, Brown, Robert Ian, Loynd, Ricky, Dewar, Robert Edward, Morton, Rachel Imogen, Wu, Qiang, Mowatt, David
• Primary Examiner(s): Vo; Huyen X.
• Application Number:
  US11/513,772
• Publication Number:
  US 20080059186A1
• Time in Patent Office: 1,727 Days
• Field of Search: 704/9, 704/251, 704/257, 704/270, 704/275, 704/1, 704/3, 704/4, 704/8, 704/10, 704/252, 704/254, 704/271
• US Class Current: 704/275
• CPC Class Codes: G10L 15/193 Formal grammars, e.g. finite state automata, context free grammars or word networks
• CPC Class Codes: G10L 2015/223 Execution procedure of a spoken command

Abstract

Intelligent speech recognition is used to provide users with the ability to utter more user friendly commands. Satisfaction is increased when a user can vocalize a subset of a formal command name and still have the intended command identified and processed. Moreover, greater accuracy in identifying a command application from a user'"'"'s utterance can be achieved by ignoring command choices associated with unlikely user utterances. An intelligent speech recognition system can identify differing acceptable verbal command phrase forms, e.g., but not limited to, complete commands, command subsequences and command subsets, for different commands supported by the system. Subset blocking words are identified for assistance in reducing the ambiguity in matching user verbal command phrases with valid commands supported by the intelligent speech recognition system.

Claims

1. A method for identifying verbal phrases, the method comprising:

defining, using a processing unit, a context free grammar for an application command of an application, the context free grammar identifying at least one valid subset for the application command, the defining comprising applying a subsequence option to a first portion of the application command to generate a first portion of a valid subset for the first portion of the application command and applying an ordered subset option to a second portion, different than the first portion, of the same application command to generate a second portion of the valid subset for the second portion of the application command, the subsequence option specifying acceptable utterances as comprising one or more command words of the application command in consecutive order with no missing interim command words, the ordered subset option specifying acceptable utterances as comprising one or more command words of the application command in order.

2. The method for identifying verbal phrases of claim 1, comprising:

generating a subset blocking word list that comprises at least one subset blocking word that is not valid within the at least one valid subset of the context free grammar.

3. The method for identifying verbal phrases of claim 2, comprising:

selectively allowing at least one subset blocking word within the subset blocking word list to be valid within the at least one valid subset of the context free grammar.

4. The method for identifying verbal phrases of claim 1, comprising:

receiving a verbal phrase from a computing device user;

generating a recognition for the verbal phrase using the context free grammar;

associating the recognition with the application command; and

providing the application command to the application for processing.

5. The method for identifying verbal phrases of claim 4, the application command being a first application command, the method comprising:

defining a context free grammar for a second application command of the application, the context free grammar for the second application command identifying at least one valid subset for the second application command;

generating a second recognition for the verbal phrase using the context free grammar for the second application command;

associating the second recognition with the second application command;

assigning a score to the first recognition based upon a likelihood of correctness of the first recognition;

assigning a score to the second recognition based upon a likelihood of correctness of the first recognition; and

providing the first application command to the application for processing if the score assigned to the first recognition is higher than the score assigned to the second recognition, else providing the second application command to the application for processing.

6. The method for identifying verbal phrases of claim 5, one context free grammar identifying the at least one valid subset for the first application command and the at least one valid subset for the second application command.

7. The method for identifying verbal phrases of claim 1, respective valid subsets of the at least one valid subset identified by the context free grammar for the application command having at least one non subset blocking word.

8. The method for identifying verbal phrases of claim 1, one or more valid subsets of the at least one valid subset for the application command being a subsequence for the application command, one of the at least one valid subset for the application command being the application command.

9. A computer storage device comprising computer executable instructions that when executed via a processor on a computer perform a method for identifying incomplete verbal commands, the method comprising:

defining a context free grammar for an application command of an application, the context free grammar identifying at least one valid subset for the application command, the defining comprising applying a subsequence option to a first portion of the application command to generate a first portion of a valid subset for the first portion of the application command and applying an ordered subset option to a second portion, different than the first portion, of the same application command to generate a second portion of the valid subset for the second portion of the application command, the subsequence option specifying acceptable utterances as comprising one or more command words of the application command in consecutive order with no missing interim command words, the ordered subset option specifying acceptable utterances as comprising one or more command words of the application command in order.

10. The method for identifying incomplete verbal commands of claim 9, comprising:

receiving an incomplete verbal command from a computing device user;

using the incomplete verbal command and the context free grammar to generate a recognition for the incomplete verbal command; and

identifying the recognition with the application command.

11. The method for identifying incomplete verbal commands of claim 10, the application command being a first application command, comprising:

defining a context free grammar for a second application command of the application, the context free grammar for the second application command identifying at least one valid subset for the second application command;

using the incomplete verbal command and the context free grammar for the second application command to generate a second recognition for the incomplete verbal command; and

requesting the computing device user identify the first application command or the second application command as the command intended by the incomplete verbal command.

12. The method for identifying incomplete verbal commands of claim 10, the application command being a first application command, comprising defining a context free grammar for a second application command, the second application command parsed into at least two application command phrases which are a subsequence of the second application command, and respective command phrases of the at least two application command phrases are defined by a context free grammar that identifies a valid subset for the application command phrase.

13. The method for identifying incomplete verbal commands of claim 12, the recognition identified with the first application command being a first recognition, and a second recognition being identified with the second application command when phrases of the incomplete verbal command match a valid subset identified for the respective application command phrases of the second application command.

14. The method for identifying incomplete verbal commands of claim 9, comprising:

generating a subset blocking word list that comprises at least one subset blocking word.

15. The method for identifying incomplete verbal commands of claim 14, the generating a subset blocking word list comprising including a word or phrase in the subset blocking word list for a punctuation mark.

16. A method for identifying verbal commands, the method comprising:

defining a first context free grammar for a first application command of a first application, the first context free grammar identifying at least one valid subset for the first application command, defining the first context free grammar comprising applying a subsequence option to a first portion of the first application command to generate a first portion of a valid subset for the first portion of the first application command and applying an orderedsubset option to a second portion, different than the first portion, of the first application command to generate a second portion of the valid subset for the second portion of the first application command; and

defining a second context free grammar for a second application command of a second application, the second context free grammar identifying at least one valid subset for the second application command, defining the second context free grammar comprising applying a subsequence option to a first portion of the second application command to generate a first portion of a valid subset for the first portion of the second application command and applying an orderedsubset option to a second portion, different than the first portion, of the second application command to generate a second portion of the valid subset for the second portion of the second application command, the subsequence option specifying acceptable utterances as comprising one or more command words of an application command in consecutive order with no missing interim command words, the ordered subset option specifying acceptable utterances as comprising one or more command words of the application command in order,

at least some of at least one of defining the first context free grammar and defining the second context free grammar implemented at least in part via a processing unit.

17. The method for identifying verbal commands of claim 16, comprising:

receiving a verbal command from a computing device user;

generating a first recognition for the verbal command using the verbal command and the first context free grammar for the first application command;

identifying the recognition with the first application command;

generating a second recognition for the verbal command using the verbal command and the second context free grammar for the second application command;

identifying the second recognition with the second application command; and

determining that the verbal command from the computing device user was intended to identify the first application command.

18. The method for identifying verbal commands of claim 17, the determining that the verbal command from the computing device user was intended to identify the first application command comprising requesting the computing device user to choose between the first application command and the second application command.

19. The method for identifying verbal commands of claim 17, comprising providing the first application command to the application for processing if a likelihood of correctness score assigned to the first recognition is higher than a likelihood of correctness score assigned to the second recognition, else providing the second application command to the application for processing.

20. The method for identifying verbal commands of claim 16, comprising:

generating a subset blocking word list that comprises at least one subset blocking word that is not valid within the at least one valid subset of the first context free grammar for the first application command and that is not valid within the at least one valid subset of the second context free grammar for the second application command.