Method for processing the output of a speech recognizer

• US 8,793,137 B1
• Filed: 07/09/2012
• Issued: 07/29/2014
• Est. Priority Date: 12/27/2007
• Status: Active Grant

• Current Assignee: Great Northern Research, LLC
• Original Assignee: Great Northern Research, LLC
• Inventors: Lagassey, Paul J., Roy, Philippe
• Primary Examiner(s): Chawan, Vijay B
• Application Number:
  US13/543,924
• Time in Patent Office: 750 Days
• Field of Search: 704/275, 704/260, 704/256.2, 704/255, 704/256, 704/270, 704/251, 704/253, 704/243, 704/254, 704/265, 704/261, 704/257, 704/267, 704/270.1, 704/268
• US Class Current: 704/275
• CPC Class Codes: F01D 5/02 Blade-carrying members, e.g. rotors rotors of non-bladed type F01D1/34; stators F01D9/00 ; selecting particular materials F01D5/28
• CPC Class Codes: F01D 5/3015 with side plates
• CPC Class Codes: F05D 2220/32 in gas turbines
• CPC Class Codes: F05D 2260/941 particularly aimed at mechanical or thermal stress reduction
• CPC Class Codes: G06F 40/205 Parsing
• CPC Class Codes: G06F 40/216 using statistical methods
• CPC Class Codes: G06F 40/289 Phrasal analysis, e.g. finite state techniques or chunking
• CPC Class Codes: G10L 15/14 using statistical models, e.g. Hidden Markov Models [HMMs] G10L15/18 takes precedence
• CPC Class Codes: G10L 15/16 using artificial neural networks
• CPC Class Codes: G10L 15/1815 Semantic context, e.g. disambiguation of the recognition hypotheses based on word meaning
• CPC Class Codes: G10L 15/19 Grammatical context, e.g. disambiguation of the recognition hypotheses based on word sequence rules
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue
• CPC Class Codes: G10L 15/285 Memory allocation or algorithm optimisation to reduce hardware requirements
• CPC Class Codes: G10L 15/34 Adaptation of a single recogniser for parallel processing, e.g. by use of multiple processors or cloud computing
• CPC Class Codes: G10L 2015/223 Execution procedure of a spoken command
• CPC Class Codes: H04L 12/282 based on user interaction within the home receiver circuitry for displaying additional information being controlled by a remote control apparatus H04N21/42204
• CPC Class Codes: H04L 2012/2847 characterised by the type of home appliance used

Abstract

A method for processing speech, comprising semantically parsing a received natural language speech input with respect to a plurality of predetermined command grammars in an automated speech processing system; determining if the parsed speech input unambiguously corresponds to a command and is sufficiently complete for reliable processing, then processing the command; if the speech input ambiguously corresponds to a single command or is not sufficiently complete for reliable processing, then prompting a user for further speech input to reduce ambiguity or increase completeness, in dependence on a relationship of previously received speech input and at least one command grammar of the plurality of predetermined command grammars, reparsing the further speech input in conjunction with previously parsed speech input, and iterating as necessary. The system also monitors abort, fail or cancel conditions in the speech input.

Claims

1. A method for processing speech, comprising:

semantically parsing a received natural language speech input with respect to a plurality of predetermined command grammars in an automated speech processing system;

determining with at least one automated processor;

if the parsed natural language speech input unambiguously corresponds to a command and comprises information that permits context-dependent statistically reliable processing with respect to completeness, then processing the command and exiting said determining;

if the received natural language speech input ambiguously corresponds to a single command or does not comprise information that permits context-dependent statistically reliable processing with respect to completeness, then prompting a user for further natural language speech input to reduce ambiguity or increase completeness, in dependence on a relationship of previously received natural language speech input and at least one command grammar of the plurality of predetermined command grammars, reparsing the further natural language speech input in conjunction with previously parsed natural language speech input, and iterating said determining; and

if an abort, fail or cancel condition is present in the natural language speech input;

generating a signal by the at least one automated processor in dependence on said determining.

2. The method according to claim 1, wherein at least one command is context sensitive, further comprising determining the context, and said determining comprises determining if the parsed natural language speech input unambiguously corresponds to the command within the determined context, and comprises information that permits context-dependent statistically reliable processing with respect to completeness.

3. The method according to claim 1, wherein the speech is processed with a statistical processor, and said determining is based on statistical probability of competing outcomes.

4. The method according to claim 1, wherein said automated speech processing system is adaptive.

5. The method according to claim 1, wherein said automated speech processing system employs both a hierarchal Markov model and a plurality of independent context free grammars.

6. The method according to claim 1, wherein said determining does not prompt a user for natural language speech input unnecessary for context-dependent statistically reliable processing.

7. The method according to claim 1, wherein said determining is responsive to at least one non-linguistic user input.

8. The method according to claim 1, wherein the natural language speech input is processed by concurrent execution in parallel in at least one virtual workspace.

9. The method according to claim 1, wherein the natural language speech input is processed by a distributed pool comprising multiple logical command processors.

10. The method according to claim 1, further comprising determining a dialect of the users, wherein the natural language speech input is selectively processed in dependence on the determined dialect of the user.

11. A system for processing speech, comprising:

a port configured to receive natural language speech input; and

at least one automated speech processor configured to;

receive the natural language speech input from the port and store it in a memory;

semantically parse the natural language speech input with respect to a plurality of predetermined command grammars;

determine if the parsed natural language speech input unambiguously corresponds to a command and comprises information that permits context-dependent statistically reliable processing with respect to completeness, then storing the command for execution and exiting the determining;

if the received natural language speech input ambiguously corresponds to a single command or does not comprise information that permits context-dependent statistically reliable processing with respect to completeness, then prompting a user for further natural language speech input to reduce ambiguity or increase completeness, in dependence on a relationship of previously received natural language speech input and at least one command grammar of the plurality of predetermined command grammars, reparsing the further natural language speech input in conjunction with previously parsed natural language speech input, and iterating the determining; and

if an abort, fail or cancel condition is detected in the natural language speech input;

generating a signal by the at least one automated processor in dependence on, said determining.

12. The system according to claim 11, wherein at least one command is context sensitive, and the automated speech processor is further configured to determine the context, and to determine if the parsed natural language speech input unambiguously corresponds to the command within the determined context, and comprises information that permits context-dependent statistically reliable processing with respect to completeness.

13. The system according to claim 11, wherein the automated processor comprises a statistical speech processor, which is configured to produce an output based on a statistical probability of competing outcomes.

14. The system according to claim 13, wherein said statistical speech processor is adaptive.

15. The system according to claim 11, wherein said automated processor employs both a hierarchal Markov model and a plurality of independent context free grammars.

16. The system according to claim 11, wherein automated processor is configured to receive at least one non-linguistic user input, and to process the natural language speech input selectively in dependence on the at least one non-linguistic user input.

17. The system according to claim 11, wherein the automated processor comprises a plurality of virtual workspaces and a plurality of processing cores.

18. The system according to claim 11, wherein the automated processor comprises a distributed pool concurrently operative logical command processors.

19. The system according to claim 11, wherein the automated processor is configured to automatically determine a dialect of a user, and to selective process the natural language speech input in dependence on the determined dialect.

20. A method for processing speech, comprising:

receiving a context of a speech input;

testing, with at least one automated processor, an output of a speech recognizer to determine a statistical reliability that indicates a command is present to be processed, and if the statistical reliability exceeds a threshold, parsing the command with respect to at least one predetermine grammar, to further determine in context sensitive manner whether the command comprises information which is unambiguous and complete with high statistical reliability;

in dependence on the testing;

processing the command if it does not comprise information which is unambiguous and complete with high statistical reliability, or

prompting a speaker to provide specific inputs to reduce the ambiguity or increase completeness, to increase the statistical reliability of the command;

or

determining whether the user seeks to abort processing of an incomplete or ambiguous command; and

selectively producing at output of the at least one automated processor in dependence on said testing.