Speech recognition application grammar modeling

• US 7,747,442 B2
• Filed: 11/21/2006
• Issued: 06/29/2010
• Est. Priority Date: 11/21/2006
• Status: Active Grant

• Current Assignee: SAP SE
• Original Assignee: SAP AG (SAP SE)
• Inventors: Doyle, Sean
• Primary Examiner(s): Wozniak; James S
• Application Number:
  US11/603,673
• Publication Number:
  US 20080120111A1
• Time in Patent Office: 1,316 Days
• Field of Search: 704243-244, 704/251, 704/270, 704/270.1, 704/275, 704/257, 379/88.01
• US Class Current: 704/270
• CPC Class Codes: G06F 8/34 Graphical or visual programming
• CPC Class Codes: G10L 15/183 using context dependencies, e.g. language models
• CPC Class Codes: G10L 15/19 Grammatical context, e.g. disambiguation of the recognition hypotheses based on word sequence rules

Abstract

The subject mater herein relates to voice applications and, more particularly, speech recognition application grammar modeling. Various embodiments provide systems, methods, and software to present one or more user interfaces through which to receive input to define and manipulate properties of a graphical listen element, wherein the properties identify one or more data sources. Some such embodiments further include building a speech recognition program grammar as a function of one or more graphical listen elements, wherein the grammar includes a representation of data retrieved from the one or more data sources.

Claims

1. A method comprising:

receiving a command to add a graphical listen element to a graphical model of a speech recognition program;

executing the received command on a processor of a computer performing the method to add the graphical listen element to the graphical model of the speech recognition program;

receiving input to manipulate properties of the graphical listen element, the properties identifying one or more data sources;

building a grammar to be used by the speech recognition program as a function of the graphical listen element, the grammar including a representation of data retrieved from the one or more data sources, wherein building the grammar includes;

for each of the one or more data sources, retrieving the data into a respective local table;

for each local table, identifying matches between data in the local table with over-ride entries; and

for each identified match with an over-ride entry, delete the matching entry in the local table and insert data from an over-ride configuration; and

storing, on a data storage device, at least one of the graphical listen element, the manipulated properties of the graphical listen element, and the grammar.

2. The method of claim 1, wherein the received input to manipulate the graphical listen element properties includes input specifying an update interval period, and wherein the grammar is rebuilt on a recurring basis in relation to the update interval period.

3. The method of claim 1, wherein the graphical listen element properties include a representation of one or more data source fields to be included in the grammar.

4. The method of claim 3, wherein the listen element properties include a representation of one or more filters to filter data source data when building the grammar.

5. The method of claim 3, wherein the graphical listen element properties include a transform rule, the transform rule identifying data to be transformed in the grammar and a form to transform the identified data into.

6. The method of claim 1, wherein the graphical listen element properties include an output representation identifying data to be output when a grammar input is received by the speech recognition program when utilizing the grammar.

7. A system comprising:

a processor;

a memory device;

a voice application modeling component stored in the memory device and executable by the processor to provide a graphical modeling environment within which voice applications are graphically modeled,

a listen element modeling component stored in the memory device and executable by the processor to provide a graphical modeling environment within which listen elements are graphically modeled, a graphically modeled listen element including properties to;

associate the listen element model with one or more of the one or more data sources, and

identify one or more data fields within the one or more data sources; and

a translation component stored in the memory device and executable by the processor to translate graphically modeled voice applications into a target metadata representation, the translation enabling a modeled voice application to be processed in a voice application runtime environment;

wherein the translation component, prior to translating a graphically modeled voice application into a target metadata representation, is executable to;

retrieve data from each of the one or more data sources into respective local tables;

for each local table, identifying matches between data in the local table with over-ride entries; and

for each identified match with an over-ride entry, delete the matching entry in the local table and insert data from an over-ride configuration.

8. The system of claim 7, wherein the translation component, when translating a graphically modeled voice application including a graphical model of a listen element:

retrieves the data of the one or more identified data fields within the one or more data sources; and

renders the retrieved data into a voice application grammar representation of the target metadata representation as a function of the graphical listen element properties.

9. The system of claim 7, wherein the properties of a graphically modeled listen element represent one or more filters to filter data of the identified data fields when processed by the translation component.

10. The system of claim 7, wherein:

the properties of the graphically modeled element include an output representation associated with the identified one or more data fields within the one or more data sources; and

the target metadata representation, when processed within the voice application runtime environment, identifies an output in response to a received input as a function of the output representation.

11. The system of claim 7, wherein an output representation identifies one or more data fields within the one or more data sources.

12. A computer-readable storage medium, including an instruction set stored thereon, which when executed by a processor of a computer causes the computer to:

present one or more user interfaces through which to receive input to define and manipulate properties of a graphical listen element, the properties identifying one or more data sources; and

build a speech recognition program grammar as a function of one or more graphical listen elements, the grammar including a representation of data retrieved from the one or more data sources, wherein building the grammar includes;

for each of the one or more data sources, retrieving the data into a respective local table;

for each local table, identifying matches between data in the local table with over-ride entries; and

for each identified match with an over-ride entry, delete the matching entry in the local table and insert data from an over-ride configuration; and

store, on a data storage device, at least one of the graphical listen element, the manipulated properties of the graphical listen element, and the grammar.

13. The computer-readable medium of claim 12, wherein the input to manipulate the graphical listen element properties includes input specifying an update interval period, wherein the grammar will be rebuilt on a recurring basis in relation to the update interval period.

14. The computer-readable medium of claim 12, wherein the graphical listen element properties allow a representation of one or more words and phrases to be included in the grammar.

15. The computer-readable medium of claim 12, wherein the graphical listen element properties allow a representation of one or more data source fields to be included in the grammar.

16. The computer-readable medium of claim 15, wherein the listen element properties allow a representation of one or more filters to filter data source data when building the grammar.

17. The computer-readable medium of claim 15, wherein the graphical listen element properties allow inclusion of a transform rule, wherein a transform rule identifies data to be transformed in the grammar and a form to transform the identified data into.

18. The computer-readable medium of claim 12, wherein the graphical listen element properties include an output representation identifying data to be output when a grammar input is received by the speech recognition program when utilizing the grammar.

19. The computer-readable medium of claim 12, wherein building the speech recognition program grammar as a function of the listen element properties includes:

converting the graphical listen element into one or more target metadata representations.

20. The computer-readable medium of claim 12, wherein the one or more data sources include an object.