METHOD AND SYSTEM FOR CAPABILITIES LEARNING

• US 20090192798A1
• Filed: 01/25/2008
• Published: 07/30/2009
• Est. Priority Date: 01/25/2008
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Basson, Sara H., Kelley, Edward E., Ramabhadran, Bhuvana, Kanevsky, Dimitri
• Granted Patent: US 8,175,882 B2
• Time in Patent Office: Days
• US Class Current: 704/270
• CPC Class Codes: G06F 40/58 Use of machine translation, e.g. for multi-lingual retrieval, for server-side translation for client devices or for real-time translation
• CPC Class Codes: G10L 15/183 using context dependencies, e.g. language models

Abstract

A method for task execution improvement, the method includes: generating a baseline model for executing a task; recording a user executing a task; comparing the baseline model to the user'"'"'s execution of the task; and providing feedback to the user based on the differences in the user'"'"'s execution and the baseline model.

Claims

1. A method for task execution improvement, the method comprising:

generating a baseline model for executing a task;

recording a user executing a task;

comparing the baseline model to the user'"'"'s execution of the task; and

providing feedback to the user based on the differences in the user'"'"'s execution and the baseline model.

2. The method of claim 1, wherein generating a baseline model further comprises:

training a machine to recognize defined and correct patterns or execution steps in a task; and

wherein the baseline model represents the defined, recognized, and accepted way to execute a task.

3. The method of claim 1, wherein the comparing is based on pattern matching.

4. The method of claim 1, wherein the task execution is offered as a pay for use service.

5. A method for accent correction, the method comprising:

training a first automatic speech recognition system (ASR1) with native speakers in a language;

training a second automatic speech recognition system (ASR2) with non-native speakers of the language;

receiving a spoken language sample from a user into ASR1 and ASR2 simultaneously;

determining whether the received spoken language sample more closely approximates the output of ASR1 or ASR2; and

wherein the level of correlation and pattern matching between the received spoken language samples and ASR1 and ASR2 serves as a feedback indicator to the user of their level of improvement.

6. The method of claim 5, wherein the output from ASR1 and ASR2 is at least one of phonetic units, words, or phrases.

7. The method of claim 5, wherein the accent correction is offered as a pay for use service.

8. The method of claim 7, wherein the pay for use service is available over the Internet.

9. A method for language translation improvement, the method comprising:

generating a baseline translation model in response to a sample text in an originating language;

generating a prediction translation model in response to the sample text with expected mistranslations in response to a user'"'"'s translation sample;

comparing the baseline translation model and the second translation model;

presenting the user with the mistranslation differences between the baseline and prediction translation models; and

wherein the baseline translation is an accurate translation of the originating language into a target language.

10. The method of claim 9, wherein the level of correlation between the baseline and prediction translation models increases as the user'"'"'s translation abilities improves.

11. The method of claim 9, wherein statistical analysis is used to create the predicted translated text with typical mistranslations.

12. The method of claim 9, wherein the comparing is based on pattern matching.

13. An article comprising one or more computer-readable storage media containing instructions that when executed by a computer enables task execution improvement;

wherein the method further comprises;

generating a baseline model for executing a task;

recording a user executing a task;

comparing the baseline model to the user'"'"'s execution of the task; and

providing feedback to the user based on the differences in the user'"'"'s execution and the baseline model.

14. The article of claim 13, wherein generating a baseline model further comprises:

training a machine to recognize defined and correct patterns or execution steps in a task; and

wherein the baseline model represents the defined, recognized, and accepted way to execute a task.

15. The article of claim 13, wherein the comparing is based on pattern matching.

16. The article of claim 15, wherein the task execution improvement is offered as a pay for use service.

17. A system for learning and task execution improvement, the system comprising:

a user dependent imitation machine that learns a user'"'"'s behaviors, actions and behaviors for carrying out a task;

a user independent machine that records defined and accepted behaviors, actions, and techniques for carrying out the task;

a comparator in electrical signal communication with the user dependent imitation machine, the user independent machine, and a learning module;

wherein the comparator determines differences in recorded actions, behaviors, and techniques between the user dependent imitation machine and the user independent machine;

wherein the learning module produces recommendations for correcting the user'"'"'s behaviors, actions, and techniques for executing the task in response to the differences determined by the comparator; and

wherein the learning module provides the recommendations as feedback to the user to improve execution of the task.

18. The system of claim 17, wherein the task is user accent correction;

wherein the user dependent imitation machine is a first automatic speech recognition system (ASR1) trained by non-native speakers in a language;

wherein the user independent machine is a second automatic speech recognition system (ASR2) trained with native speakers of the language;

wherein the system;

receives a spoken language sample from a user into ASR1 and ASR2 simultaneously;

determines whether the received spoken language sample more closely approximates the output of ASR1 or ASR2; and

wherein the level of correlation and pattern matching between the received spoken language samples and ASR1 and ASR2 serves as a feedback indicator from the learning module to the user of their level of improvement.

19. The system of claim 18, wherein the output from ASR1 and ASR2 is at least one of phonetic units, words, or phrases.

20. The system of claim 18, wherein the accent correction is offered as a pay for use service.

21. The system of claim 20, wherein the pay for use service is available over the Internet.