Method and apparatus for building asset based natural language call routing application with limited resources

• US 20080010280A1
• Filed: 06/16/2006
• Published: 01/10/2008
• Est. Priority Date: 06/16/2006
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: International Business Machines Corporation
• Inventors: Stewart, Osamuyimen Thompson, Jan, Ea-Ee, Li, Xiang, Lubensky, David
• Granted Patent: US 8,370,127 B2
• Time in Patent Office: Days
• US Class Current: 707/7
• CPC Class Codes: G10L 15/063 Training
• CPC Class Codes: G10L 15/1822 Parsing for meaning understanding
• CPC Class Codes: G10L 15/183 using context dependencies, e.g. language models

Abstract

A method of processing limited natural language data to automatically develop an optimal feature set, bypassing the standard Wizard of OZ (WOZ) approach is provided. The method provides for building natural language understanding models or for processing existing data from other domains, such as the Internet, for domain-specific adaptation through the use of an optimal feature set. Consequently, when the optimal feature set is passed on to any engine, the optimal feature set produces robust models that can be used for natural language call routing.

Claims

1. A computer-implemented method for building data used by an understanding model of a natural language call routing application, the computer-implemented method comprising:

providing a plurality of topic descriptions, wherein each topic description of the plurality of topic descriptions describes a meaning of a topic of a plurality of topics;

providing training data, wherein the training data is based on the plurality of topic descriptions;

identifying keywords in the training data; and

creating an optimal feature set, wherein the optimal feature set is based on the keywords.

2. The computer implemented method of claim 1, further comprising:

identifying filler words in the training data; and

removing the filler words from the training data.

3. The computer implemented method of claim 2, further comprising:

maintaining a list of filter words.

4. The computer implemented method of claim 1, wherein the plurality of topics is based on a voice user interface specification.

5. The computer implemented method of claim 1, wherein each topic of the plurality of topics corresponds to a unit of action, wherein the unit of action defines what a user may say and a response to what the user may say.

6. The computer implemented method of claim 1, further comprising:

creating an initial topic classification model based on the optimal feature set.

7. The computer implemented method of claim 6, further comprising:

tagging collected data using the initial topic classification model, wherein the tagging is based upon a Semantic meaning of the collected data

8. The computer implemented method of claim 1, wherein the optimal feature set comprises at least one of keywords and a combination of keywords.

9. A computer program product comprising a computer usable medium including computer usable program code for building data used by an understanding model of a natural language call routing application, the computer-program product comprising:

computer usable program code for providing a plurality of topic descriptions, wherein each topic description of the plurality of topic descriptions describes a meaning of a topic of a plurality of topics;

computer usable program code for providing training data, wherein the training data is based on the plurality of topic descriptions;

computer usable program code for identifying keywords in the training data; and

computer usable program code for creating an optimal feature set, wherein the optimal feature set is based on the keywords.

10. The computer program product of claim 9, further comprising:

computer usable program code for identifying filler words in the training data; and

computer usable program code for removing the filler words from the training data.

11. The computer program product of claim 10, farther comprising:

computer usable program code for maintaining a list of filler words.

12. The computer program product of claim 9, wherein the plurality of topics are based on a voice user interface specification.

13. The computer program product of claim 9, wherein each topic of the plurality of topics corresponds to a unit of action, wherein the unit of action defines what a user may say and a response to what the user may say.

14. The computer program product of claim 9, further comprising:

computer usable program code for creating an initial topic classification model based on the optimal feature set.

15. The computer program product of claim 14, further comprising:

computer usable program code for tang collected data, wherein the tagging is based upon a semantic meaning within the collected data.

16. The computer program product of claim 9, wherein the optimal feature set comprises at least one of keywords and a combination of keywords.

17. A data processing system for building data used by an understanding model of a natural language call routing application, the data processing system comprising:

a storage device, wherein the storage device stores computer usable program code; and

a processor, wherein the processor executes the computer usable program code to provide a plurality of topic descriptions, wherein each topic description of he plurality of topic descriptions describes a meaning of a topic of the plurality of topics;

provide training data, wherein the training data is based on the plurality of topic descriptions;

identify keywords in the training data; and

create an optimal feature sets wherein the optimal feature set is based on the keywords.

18. The data processing system of claim 17, wherein the processor further executes the computer usable program code to identify filler words in the training data;

and remove the filler words from the training data.

19. The data processing system of claim 18, wherein the processor further executes the computer usable program code to maintain a list of filler words.

20. The data processing system of claim 17, wherein the processor further executes the computer usable program code to create an initial topic classification model based on the optimal feature set.