US 10,832,161 B2Granted

Method and system of processing data for training a target domain classifier

AssigneeConduent Business Services, LLC

InventorsHimanshu Sharad Bhatt, Raghuram Krishnapuram

Life of this patent

Estimating the term…

Loading patent…

Abstract

The disclosed embodiments illustrate method and system of processing data by a computing device for training a target domain classifier. The method includes extracting one or more first features and one or more second features from a first target instance, associated with a target domain. The method further includes predicting a first label for the received first target instance based on the one or more first features by utilizing a trained first classifier associated with a set of labeled source instances, wherein the predicted first label is assigned to the first target instance when a first score of the predicted first label exceeds a first pre-specified threshold. Further, the method includes updating a set of labeled target instances associated with the target domain based on the labeled first target instance, wherein the updated set of labeled target instances is utilized to train the target domain classifier.

Claims18

Dependent claims

A method of processing data by a computing device for training a target domain classifier, the method comprising:

extracting, by one or more processors in a computing device, one or more first features and one or more second features from a first target instance, associated with a target domain, received from another computing device;
training, by the one or more processors, a first classifier based on one or more labeled source features associated with a set of labeled source instances and/or one or more labeled target features associated with a set of labeled target instances;
predicting, by the one or more processors in the computing device, a first label for the received first target instance based on the one or more first features by utilizing the trained first classifier, wherein the predicted first label is determined as:
Y
=
arg
⁢
⁢
max
⁢
⁢
P
⁡
(
Y
=
y
k
)
y
∈
Y
⁢
∏
i
⁢
P
⁡
(
x
i
l
|
Y
=
y
k
)
where:
Y represents the predicted first label,
xil represents an ith feature in the one or more first features of the first target instance, and
yk represents a label associated with a feature same as xil in the source feature log and/or the target feature log with a maximum count;
determining a first score based on a ratio between a count of the one or more first features and a total count of the one or more first features and the one or more second features;
assigning the predicted first label to the first target instance when the first score of the predicted first label exceeds a first pre-specified threshold;
assigning the predicted first label is assigned to the one or more second features;
updating, by the one or more processors in the computing device, the set of labeled target instances associated with the target domain based on the labeled first target instance, wherein the updated set of labeled target instances is utilized to train the target domain classifier when a cumulative value associated with the set of labeled target instances exceeds a pre-specified threshold value;
and
predicting, by the one or more processors in the computing device, a second label for the received first target instance based on the one or more first features and the one or more second features by utilizing a trained second classifier associated with a source domain when the first score of the predicted first label is less than the first pre-specified threshold, wherein the predicted second label is assigned to the first target instance when a second score associated with the predicted second label exceeds a second pre-specified threshold.

45 more elements

A system for processing data by a computing device for training a target domain classifier, the system comprising:

one or more processors in a computing device configured to:
extract one or more first features and one or more second features from a first target instance, associated with a target domain, received from another computing device;
train a first classifier based on one or more labeled source features associated with a set of labeled source instances and/or one or more labeled target features associated with a set of labeled target instances;
predict a first label for the received first target instance based on the one or more first features by utilizing a trained first classifier associated with a set of labeled source instances, wherein the predicted first label is determined as:
Y
=
arg
⁢
⁢
max
⁢
⁢
P
⁡
(
Y
=
y
k
)
y
∈
Y
⁢
∏
i
⁢
P
⁡
(
x
i
l
|
Y
=
y
k
)
where:
Y represents the predicted first label,
xil represents an ith feature in the one or more first features of the first target instance, and
yk represents a label associated with a feature same as xil in the source feature log and/or the target feature log with a maximum count;
determine a first score based on a ratio between a count of the one or more first features and a total count of the one or more first features and the one or more second features;
assign the predicted first label to the first target instance when the first score of the predicted first label exceeds a first pre-specified threshold:
assign the predicted first label to the one or more second features;
update a set of labeled target instances associated with the target domain based on the labeled first target instance, wherein the updated set of labeled target instances is utilized to train the target domain classifier when a cumulative value associated with the set of labeled target instances exceeds a pre-specified threshold value;
and
predict, by the one or more processors in the computing device, a second label for the received first target instance based on the one or more first features and the one or more second features by utilizing a trained second classifier associated with a source domain when the first score of the predicted first label is less than the first pre-specified threshold, wherein the predicted second label is assigned to the first target instance when a second score associated with the predicted second label exceeds a second pre-specified threshold.

46 more elements

A computer program product for use with a computer, the computer program product comprising a non-transitory computer readable medium, wherein the non-transitory computer readable medium stores a computer program code of data processing for training a target domain classifier, wherein the computer program code is executable by one or more processors in a computing device to:

extract one or more first features and one or more second features from a first target instance, associated with a target domain, received from another computing device;
train a first classifier based on one or more labeled source features associated with a set of labeled source instances and/or one or more labeled target features associated with a set of labeled target instances;
predict a first label for the received first target instance based on the one or more first features by utilizing a trained first classifier associated with a set of labeled source instances, wherein the predicted first label is determined as:
Y
=
arg
⁢
⁢
max
⁢
⁢
P
⁡
(
Y
=
y
k
)
y
∈
Y
⁢
∏
i
⁢
P
⁡
(
x
i
l
|
Y
=
y
k
)
where:
Y represents the predicted first label,
xil represents an ith feature in the one or more first features of the first target instance, and
yk represents a label associated with a feature same as xil in the source feature log and/or the target feature log with a maximum count;
determine a first score based on a ratio between a count of the one or more first features and a total count of the one or more first features and the one or more second features;
assign the predicted first label to the first target instance when the first score of the predicted first label exceeds a first pre-specified threshold:
assign the predicted first label to the one or more second features;
update a set of labeled target instances associated with the target domain based on the labeled first target instance, wherein the updated set of labeled target instances is utilized to train the target domain classifier when a cumulative value associated with the set of labeled target instances exceeds a pre-specified threshold value;
and
predict, by the one or more processors in the computing device, a second label for the received first target instance based on the one or more first features and the one or more second features by utilizing a trained second classifier associated with a source domain when the first score of the predicted first label is less than the first pre-specified threshold, wherein the predicted second label is assigned to the first target instance when a second score associated with the predicted second label exceeds a second pre-specified threshold.

45 more elements

Dependencies are inferred from the claim text.