Method for transforming data elements within a classification system based in part on input from a human annotator or expert
View Patent ↗A method is provided for transforming data elements within a classification system based in part on input from a human annotator or expert. A first concept evolution model as a training set is composed from a first set of selectively determinable annotations and the first concept evolution model. A trained model is generated after training a learning algorithm with the training set and the concept evolution model. A confidence factor is computed that a predicted annotation is accurately identified. A selected element instance and a corresponding suggested annotation are identified to have a low confidence factor. The classifying of the applied annotation is adjusted where a second concept evolution model is composed for more accurate classifying of the data item.
1. A method for evolving an annotating model for classifying a document or a data item therein, comprising:
composing a first concept evolution model as a training set comprised of a first set of selectively determinable class labels of element instances within the document that are detectable within the document to produce a result of predicting class labels to be assigned to unlabeled element instances and the first concept evolution model;
training a learning algorithm with the training set and the concept evolution model to generate a trained model wherein the learning algorithm comprises a global approach to reshape a list of the classes and adjusts the set of features, or wherein the learning algorithm comprises a local approach that creates a local model of one or few events, the definition set of classes remains unchanged, and the training set can be extended with new examples;
using the trained model to predict class labels for unlabeled element instances within the document;
computing a confidence factor for a predicted class label is accurately predicted for unlabeled elements;
identifying an unlabeled element instance within the document with a corresponding suggested annotation having a confidence factor less than zero; and
adjusting the classifying of the unlabeled element instance wherein a second concept evolution model is composed for more accurate classifying of the document, and wherein the composing and applying are executed by a designer of the annotating model and the computing is machine implemented.
2. The method of claim 1 wherein the composing comprises associating a class with detectable annotations.
3. The method of claim 2 wherein the computing comprises determining a probability that a detected annotation corresponds to a class, and when the probabilities for all classes correspond to the confidence factor satisfying the predetermined condition of the uncertainty, suggesting annotating of the class to an annotator or expert.
4. The method of claim 3 wherein the adjusting comprises the local approach concept evaluation comprising associating a local model for each evolution event including a concept evolution command.
5. The method of claim 4 wherein the associating a local model comprises corresponding an event model to an internal mode of a concept evolution DAG.
6. The method of claim 3 wherein the adjusting comprises a global approach concept evolution including associating a global model for a most recent changing of the associate features for the predicted class comprising issuing of a concept evolution command by the annotator or expert.
7. The method of claim 6 wherein the associating a global model comprises changing the set of classes in accordance with the issued concept evolution command and removing annotations for the data items that are obsolete from the changing.
8. The method according to claim 1 , wherein the confidence factor is calculated using the formula:
conf
me
(
x
)
=
∑
i
P
(
y
i
|
x
)
log
P
(
y
i
|
x
)
.
9. The method according to claim 1 , wherein the confidence factor is normalized using the formula:
conf
me
n
(
x
)
=
1
log
Y
∑
i
P
(
y
i
|
x
)
log
P
(
y
i
|
x
)
.