Double-barreled question predictor and correction

1. A computer-implemented method, comprising:

gathering data samples into a data set;

correcting for imbalance in the data set to produce a corrected data set by applying active learning to the data set to increase a number of double-barreled question data samples occurring in the data set;

selecting an optimal machine learning model for the corrected data set;

training the optimal machine learning model using the corrected data set;

operating the optimal machine learning model on new data to produce a prediction result; and

generating a visual representation of at least one prediction results.

2. The computer-implemented method as claimed in claim 1 , wherein correcting for imbalance in the data set by applying active learning comprises;

determining a sampling score for each sample;

identifying a sample having a highest sampling score;

acquiring a label for the sample having the highest uncertainty score;

adding the sample to the training dataset; and

repeating the determining, identifying, acquiring and adding to produce the corrected data set.

3. The computer-implemented method as claimed in claim 2 , wherein determining the sampling score for each sample comprises determining an uncertainty measure for each sample, the uncertainty measure being a measure of whether a particular sample resides in a particular class.

4. The computer-implemented method as claimed in claim 3 , wherein determining the uncertainty measure comprises determining the uncertainty measure using at least two classifiers, and combining the results from the at least two classifiers.

5. The computer-implemented method as claimed in claim 1 , wherein selecting the optimal machine learning model comprises:

generating at least two vector representations for an input text portion of each sample from at least two different methods;

combining the at least two vector representations in different combinations;

using cross validation to identify a best combination of the vector representations;

using the best combination to represent the input text portion of the sample in the training dataset; and

using the training dataset to train the machine learning model.

6. The computer-implemented method as claimed in claim 5 , wherein generating at least two vector representations comprises generating at least one word-level vector representation and at least one sentence-level vector representation.

7. The computer-implemented method as claimed in claim 5 , wherein combining the at least two vector representations comprises applying a mathematical operator to the at least two vector representations to produce the combinations.

8. The computer-implemented method as claimed in claim 1 , wherein generating a visualization of the prediction result comprises displaying a value for each token in an input text portion of each data sample.

9. The computer-implemented method as claimed in claim 1 , further comprising:

receiving at least one data sample from a user as the new data;

operating the optimal machine learning model on the new data to produce a prediction result, wherein the prediction result is whether the data sample from the user is a double-barreled question; and

when the question is a double-barreled question, identifying the question as a double-barreled question to the user.

10. The computer-implemented method as claimed in claim 1 , further comprising retraining the optimal machine learning model based upon one of either performance of the model or a periodic schedule.

11. The computer-implemented method as claimed in claim 10 , wherein performance of the model is based upon the visualization of the prediction result.

12. A computer-implemented method to correct for imbalance in a data set using active learning, comprising:

determining a sampling score for each sample in the data set using a committee of at least two classifiers to each provide a sampling score and then combining the sample scores by using an absolute value of a sum of a number of models to choose a particular class for each sample divided by the number of overall classifiers to produce the sampling score;

identifying a sample having a highest sampling score;

acquiring a label for the sample having the highest sampling score;

adding the sample to a training data set; and

repeating the determining, identifying, acquiring, and adding to produce the training data set.

13. The computer-implemented method as claimed in claim 12 , wherein determining the sampling score for each sample comprises determining an uncertainty measure for each sample, the uncertainty measure being a measure of whether a particular sample resides in a particular class.

14. A computer-implemented method to select an optimal machine learning model comprises:

generating at least two vector representations of a data sample comprised of an input pair comprising question text and a corresponding answer option text in a training set from at least two different methods of generating word vector representations;

combining the at least two vector representations in different combinations using different methods of combining;

using cross validation to identify a best method of combination of the vector representations;

using the best combination to represent the input pair of the sample in the training dataset; and

using the training data set to train the machine learning model.

15. The computer-implemented method as claimed in claim 14 , wherein generating at least two vector representations comprises generating at least one word-level vector representation and at least one sentence-level vector representation.

16. The computer-implemented method as claimed in claim 14 , wherein combining the at least two vector representations comprises applying a mathematical operator to the at least two vector representations to produce the combinations.