Model development authoring, generation and execution based on data and processor dependencies

• US 8,086,455 B2
• Filed: 01/09/2008
• Issued: 12/27/2011
• Est. Priority Date: 01/09/2008
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Gong, Yifan, Tian, Ye
• Primary Examiner(s): Chawan, Vijay B
• Application Number:
  US11/971,897
• Publication Number:
  US 20090177471A1
• Time in Patent Office: 1,448 Days
• Field of Search: 704/235, 704/242, 704/251, 704/249, 704/231, 704/250, 704/243, 704/244, 704/256.2, 704/256, 718/104, 718/106, 705/7.36, 703/13
• US Class Current: 704/242
• CPC Class Codes: G06F 18/40 Software arrangements specially adapted for pattern recognition, e.g. user interfaces or toolboxes therefor
• CPC Class Codes: G06F 2209/485 Resource constraint
• CPC Class Codes: G06Q 10/0637 Strategic management or analysis, e.g. setting a goal or target of an organisation; Planning actions based on goals; Analysis or evaluation of effectiveness of goals
• CPC Class Codes: G10L 15/28 Constructional details of speech recognition systems

Abstract

A recognition (e.g., speech, handwriting, etc.) model build process that is declarative and data-dependence-based. Process steps are defined in a declarative language as individual processors having input/output data relationships and data dependencies of predecessors and subsequent process steps. A compiler is utilized to generate the model building sequence. The compiler uses the input data and output data files of each model build processor to determine the sequence of model building and automatically orders the processing steps based on the declared input/output relationship (the user does not need to determine the order of execution). The compiler also automatically detects ill-defined processes, including cyclic definition and data being produced by more than one action. The user can add, change and/or modify a process by editing a declaration file, and rerunning the compiler, thereby a new process is automatically generated.

Claims

1. A computing device for implementing a recognition model development system, comprising:

a memory; and

a processor coupled to the memory, the processor executing a plurality of components associated with the recognition model development system, wherein the components include;

a programming component for defining recognition model build steps as declarations of processors and, input data and output data of the processors as dependencies of predecessor and subsequent process steps to form a declarative and data-dependent model build process; and

an execution component for ordering and executing the steps based on the declarations to generate recognition model actions.

2. The computing device of claim 1, wherein the steps are arranged to have the input data available before execution.

3. The computing device of claim 1, further comprising a compiler for generating an initial model development graph having processors and pieces of data as nodes and dependencies as edges.

4. The computing device of claim 3, further comprising a validation component for validating the initial model graph for data dependencies for the processors.

5. The computing device of claim 1, wherein the execution component includes a generator component for generating an ordered list of the processors, each of the processors having input data available for execution.

6. The computing device of claim 5, further comprising an interpreter component for examining dates of the data and determining next processors for execution based on the dates.

7. A computing device for implementing a recognition model development system, comprising:

a memory; and

a processor coupled to the memory, the processor executing a plurality of components associated with the recognition model development system, wherein the components include;

a programming component for defining speech recognition model build steps according to a declarative language, each step defined as a processor having a data dependency with predecessor and subsequent processors and an input/output mapping to form a declarative and data-dependent model build process;

a generator component for generating an ordered list of processors where input data for each processor is available; and

an execution component for executing the steps based on the ordered list of processors and generating speech recognition model actions.

8. The computing device of claim 7, further comprising a compiler for generating an initial model development graph having processors as nodes and dependencies as edges, and a validation component for validating the initial model graph for data dependencies for the processors.

9. The computing device of claim 8, further comprising a detection component for detecting if the graph is cyclic.

10. The computing device of claim 7, further comprising a task wrapper for wrapping tools, executables and scripts into a compatible interface for activation by the execution component.

11. A computer-readable memory device with instructions stored thereon for executing a method of developing recognition models, the method comprising:

authoring recognition model build steps declaratively as a task description, the steps each including data and a processor for obtaining the data, and having a data dependency with predecessor and subsequent processor;

generating a model development process graph of processor input data and output data based on the task description;

generating an ordered list of the steps for execution; and

executing the ordered list of steps to output recognition model actions.

12. The computer-readable memory device of claim 11, wherein the recognition model is a speech recognition model.

13. The computer-readable memory device of claim 11, wherein the method further comprises generating a predecessor step for each of the steps based on the task description.

14. The computer-readable memory device of claim 11, wherein the method further comprises validating consistency of each step based on the task description.

15. The computer-readable memory device of claim 11, wherein the method further comprises verifying that the input data for each processor is available for processor execution.

16. The computer-readable memory device of claim 11, wherein the method further comprises generating a list of potential output data for one or more processors based on the input data.

17. The computer-readable memory device of claim 11, wherein the method further comprises generating a list of potential input data for one or more processors based on the output data.

18. The computer-readable memory device of claim 11, wherein the method further comprises automatically programming order of execution of the steps based on data dependencies defined in the task description.

19. The computer-readable memory device of claim 11, wherein the method further comprises implementing a development process by changing declarations for a predefined set of steps in the task description and executing the predefined set of the steps.

20. The computer-readable memory device of claim 11, wherein the method further comprises adding one or more new processors in the form of new declarations in the task description without recoding or retesting.