DISTANCE METRICS FOR UNIVERSAL PATTERN PROCESSING TASKS

• US 20090259471A1
• Filed: 04/11/2008
• Published: 10/15/2009
• Est. Priority Date: 04/12/2007
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Sainath, Tara N., KANEVSKY, Dimitri, Nahamoo, David
• Granted Patent: US 8,856,002 B2
• Time in Patent Office: Days
• US Class Current: 704/260
• CPC Class Codes: G10L 15/063 Training
• CPC Class Codes: G10L 17/04 Training, enrolment or model building

Abstract

A universal pattern processing system receives input data and produces output patterns that are best associated with said data. The system uses input means receiving and processing input data, a universal pattern decoder means transforming models using the input data and associating output patterns with original models that are changed least during transforming, and output means outputting best associated patterns chosen by a pattern decoder means.

Claims

1. A universal pattern processing system receiving input data and producing output patterns that are best associated with said data, comprising:

input means receiving and processing input data;

universal pattern decoder means comprising transforming means, said decoder transforming models using the input data and associating output patterns with original models that are changed least during the transforming; and

output means outputting best associated patterns chosen by the universal pattern decoder.

2. The system as in claim 1, where input data comprises at least one of the following:

electronic signals, optical text, audio, optical images, video, binary data, analog data, and biometrics.

3. The system as in claim 1, where the universal pattern decoder comprises at least one of the following:

a speech recognizer, an image recognizer, a segmentation system, a machine translation, a search engine, a text-to-speech, a pattern recognizer, and a language processor.

4. The system as in claim 1, wherein the transformation means comprises at least one of the following:

Maximum Likelihood training, Maximum Entropy training, EM training, Maximum Mutual Information training, Baum-Welch training, Extended Baum-Welch training, Bayesian network training, supervised training, unsupervised training, and Hidden Markov Model training.

5. The system as in claim 1, wherein a measuring of a change during a transformation comprises at least one of the following metrics:

a) one or more differences scores of transformed and original models;

b) one or more of slopes through a score curve of model transformations;

c) one or more tangents to a score curve of model transformations;

d) one or more tangent hyperplanes to a score manifold of model transformations;

e) a vector support machine separation metric for one or more objects in a)-d);

f) a Hidden Markov Models metric that is accumulated from local Hidden Markov Model metrics for Hidden-Markov-Model states.

6. A method of universal pattern processing in a universal pattern processor that receives input data and produces output patterns that are best associated with this data, said method comprising:

receiving and processing input data;

transforming universal pattern processing models using the input data;

associating output patterns with original models that changed least during the transformation process; and

outputting best associated patterns.

7. The method as in claim 6, wherein the input data processing comprises at least one of the following:

signal processing, textual processing, audio processing, image processing, video processing, binary data processing, analog data processing, biometrics sensor processing, and language processing.

8. The method as in claim 6, wherein the universal pattern decoder process comprises at least one of the following:

speech recognition process, image recognition process, segmentation process, machine translation, search, text-to-speech process, pattern recognition process, and summarization, simplification.

9. The method as in claim 6, wherein the transformation process comprises at least one of the following:

Maximum Likelihood training, Maximum Entropy training, EM training, Maximum Mutual Information training, Baum-Welch training, Extended Baum-Welch training, Bayesian network training, supervised training, and unsupervised training.

10. The method as in claim 6, wherein measuring a change of model process during a transformation comprises at least one of the following:

a) subtracting scores of transformed models from scores of original models;

b) computing slopes through a score curve of model transformations;

c) computing tangents to a score curve of model transformations;

d) computing tangent hyperplanes to a score manifold of model transformations;

e) computing support machine separation metric for one or more objects in a)-d);

f) computing a Hidden Markov Models metric that is accumulated from local Hidden Markov Model metrics for Hidden Markov Model states.

11. A computer-readable medium on which are encoded machine-readable instructions which, when executed, cause a computer to perform the method of claim 6.

12. A digital computer comprising the computer-readable medium of claim 11.