Using strong data types to express speech recognition grammars in software programs

• US 7,460,996 B2
• Filed: 06/23/2005
• Issued: 12/02/2008
• Est. Priority Date: 06/23/2005
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Peyroux, Jean-Francois
• Primary Examiner(s): MCFADDEN, SUSAN IRIS
• Application Number:
  US11/159,648
• Publication Number:
  US 20060293894A1
• Time in Patent Office: 1,258 Days
• Field of Search: 704/257
• US Class Current: 704/257
• CPC Class Codes: G06F 40/253 Grammatical analysis; Style critique
• CPC Class Codes: G06F 40/30 Semantic analysis
• CPC Class Codes: G10L 15/1815 Semantic context, e.g. disambiguation of the recognition hypotheses based on word meaning
• CPC Class Codes: G10L 15/183 using context dependencies, e.g. language models
• CPC Class Codes: G10L 2015/228 of application context

Abstract

In one embodiment, a system for interpretation of natural language inputs to natural language enabled applications has a grammar, a source file, and a speech recognizer. The grammar has a set of data types and semantic code defining rules about the set of data types. The source file includes computer readable code adapted to access the grammar by instantiating selected data types of the set of data types. The speech recognizer is adapted to receive natural language input and to map the natural language input to the selected data types.

Claims

1. A system for interpretation of natural language inputs to natural language enabled applications comprising:

a grammar comprising a set of object-oriented data types and semantic code defining rules about the set of data types, wherein the set of object-oriented data types includes a plurality of classes;

a source file comprised of computer readable code to access the grammar by instantiating selected data types of the set of object-oriented data types with one or more parameters to produce strongly-typed grammars with associated parameters; and

a speech recognizer to receive natural language input and to map the natural language input to strongly-typed grammars according to the associated parameters, and wherein the speech recognizer generates a recognition output based on the mapping.

2. The system of claim 1 further comprising:

a grammar compiler to receive the grammar and to separate the set of data types from the semantic code, wherein the grammar compiler generates a strongly typed grammar based on the grammar.

3. The system of claim 2 wherein the strongly typed grammar comprises a state machine.

4. The system of claim 1 wherein the selected data types are instantiated with the one or more parameters at runtime.

5. The system of claim 1 further comprising:

a code compiler to receive a source code file and the source file and to compile the source code and the source file into an intermediate language code.

6. The system of claim 5 wherein the intermediate language code comprises computer readable code and associated debug information.

7. A system for natural language input recognition comprising:

an input device configured to receive a natural language input;

a grammar compiler configured to generate a strongly typed grammar comprising a dynamic link library defining a set of object-oriented data types representative of elements of a natural language, wherein the dynamic link library comprises a state machine containing multiple linguistic versions of a grammar as categories and an intermediate language code comprising semantic processing information and debugging information;

a recognition component configured to convert the natural language input into text using the state machine to categorize the text based on types defined within the dynamic link library of the strongly typed grammar and to generate a recognition output based on the text, the recognition output comprising categorized words of the natural language input; and

an interpretation component configured to receive the categorized words and to identify semantics of the natural language input based on the categorized words.

8. The system of claim 7 wherein the semantic processing information defines meanings of types within the strongly typed grammar.

9. The system of claim 8 wherein the interpretation component utilizes the semantic processing information to identify semantics of the natural language input based on the grammar.

10. The system of claim 7 wherein the recognition component is configured to convert the natural language input into text specified in the grammar.

11. The system of claim 7 wherein the interpretation component comprises one or more interpretation components, each interpretation component configured to generate a possible interpretation of the natural language input, the system further comprising:

an integration component configured to integrate a respective one or more possible interpretations of the natural language input generated by the one or more interpretation components, wherein the integration component includes a processor configured to select types from the identified semantics that can be used by a client device, wherein types that cannot be used by the client device are discarded and types that can be used by the client device are combined into an integrated output grammar object with fields and properties, wherein the integrated output grammar object is configured for use by a client application on the client device.

12. The system of claim 7 wherein the natural language input comprises speech.

13. The system of claim 7 wherein the natural language input comprises hand-writing.

14. The system of claim 7 further comprising:

application code defined within a grammar file containing the strongly typed grammar and within one or more dedicated source code files, wherein the grammar compiler generates the strongly typed grammar from the one or more dedicated source code files.

15. A method for accessing grammars for natural language applications comprising:

providing one or more grammars, each grammar comprising a set of object-oriented data types representative of a natural language;

defining processes for operating on a natural language input in one or more source code files;

instantiating selected data types of the set of object-oriented data types with one or more parameters to produce strongly-typed grammars with associated parameters at runtime, the selected types corresponding to a natural language input;

receiving the natural language input and mapping the natural language input to the strongly-typed grammars according to the associated parameters; and

generating a recognition output based on the mapping.

16. The method of claim 15 further comprising:

compiling the grammar into a state machine using a grammar compiler; and

providing the state machine and semantic code associated with the selected types to a speech recognizer.

17. The method of claim 15 wherein the step of instantiating further comprises:

creating one class for each public grammar rule within the one or more grammars using a grammar compiler.

18. The method of claim 15 wherein each grammar further comprises semantic code associated with the set of object-oriented data types.

19. The method of claim 15 wherein the selected data types inherit properties from the grammar based on a data type definition.