Text classifier training

1. A method, in a data processing system, for text classifier training, the method comprising:

training a plurality of text classifier machine learning models based on a plurality of replacement method maps;

classifying, by a text classifier component configured with each of the plurality of the trained text classifier machine learning models, a test set of input texts to form a plurality of sets of class determinations, wherein each set of class determinations corresponds to a respective one of the plurality of replacement method maps;

determining an accuracy of each of the plurality of sets of class determinations;

configuring a domain terms replacer component with the replacement method map corresponding to a set of class determinations having a highest accuracy;

receiving, by the data processing system, a training set of text and class specification pairs to be used as a ground truth for training a text classifier machine learning model for a text classifier;

identifying, by the domain terms selector component executing on a processor of the data processing system, at least one domain term in the texts of the training set;

replacing, by a domain terms replacer component executing on a processor of the data processing system, the at least one identified domain term in the texts of the training set with a corresponding replacement term to form a revised set of text and class specification pairs; and

training, by a text classifier trainer component executing on a processor of the data processing system, the text classifier machine learning model using the revised set to form a trained text classifier machine learning model,

wherein replacing the at least one identified domain term comprises, for each given identified domain term, determining a domain term type of the given identified domain term and replacing the given identified domain term based on selecting a replacement method associated with the domain term type by a replacement method map, wherein the trained text classifier machine learning model classifies an input text into one of a plurality of text classes selected from the group consisting of: sentiment, factoid, descriptive, and reasoned.

2. The method of claim 1 , wherein for at least one identified domain term, the corresponding replacement term is an empty string.

3. The method of claim 1 , wherein for at least one identified domain term, the corresponding replacement term is a different domain term substring comprising a description of what the identified domain term contains.

4. The method of claim 3 , wherein the different domain term substring identities an entity canonical name, an entity type, a relation, or a negative relation.

5. The method of claim 1 , wherein the replacement method is identified from a set consisting of: replacing the given identified domain term with an empty string, replacing the given identified domain term with a different domain term substring, or not replacing the given identified domain term.

6. A computer program product comprising a non-transitory computer readable storage medium having a computer readable program stored therein, wherein the computer readable program, when executed on a computing device, causes the computing device to implement a domain terms selector component, a domain terms replacer component, and a text classifier trainer component, wherein the computer readable program causes the computing device to:

train a plurality of text classifier machine learning models based on a plurality of replacement method maps;

classify, by the text classifier component configured with each of the plurality of the trained text classifier machine learning models, a test set of input texts to form a plurality of sets of class determinations, wherein each set of class determinations corresponds to a respective one of the plurality of replacement method maps;

determine an accuracy of each of the plurality of sets of class determinations;

configure the domain terms replacer component with the replacement method map corresponding to a set of class determinations having a highest accuracy;

receive, by the data processing system, a training set of text and class specification pairs to be used as a ground truth for training a text classifier machine learning model for a text classifier;

identify, by the domain terms selector component executing on a processor of the data processing system, at least one domain term in the texts of the training set;

replace, by the domain terms replacer component executing on a processor of the data processing system, the at least one identified domain term in the texts of the training set with a corresponding replacement term to form a revised set of text and class specification pairs; and

train, by the text classifier trainer component executing on a processor of the data processing system, the text classifier machine learning model using the revised set to form a trained text classifier machine learning model,

wherein replacing the at least one identified domain term comprises, for each given identified domain term, determining a domain term type of the given identified domain term and replacing the given identified domain term based on selecting a replacement method associated with the domain term type by a replacement method map, wherein the trained text classifier machine learning model classifies an input text into one of a plurality of text classes selected from the group consisting of: sentiment, factoid, descriptive, and reasoned.

7. The computer program product of claim 6 , wherein for at least one identified domain term, the corresponding replacement term is an empty string.

8. The computer program product of claim 6 , wherein for at least one identified domain term, the corresponding replacement term is a different domain term substring comprising a description of what the identified domain term contains.

9. The computer program product of claim 8 , wherein the different domain term substring identifies an entity canonical name, an entity type, a relation, or a negative relation.

10. The computer program product of claim 6 , wherein the replacement method is identified from a set consisting of: replacing the given identified domain term with an empty string, replacing the given identified domain term with a different domain term substring, or not replacing the given identified domain term.

11. An apparatus comprising:

a processor; and

a memory coupled to the processor, wherein the memory comprises instructions which, when executed by the processor, cause the processor to implement a domain terms selector component, a domain terms replacer component, and a text classifier trainer component, wherein the instructions cause the processor to:

train a plurality of text classifier machine learning models based on a plurality of replacement method maps;

classify, by the text classifier component configured with each of the plurality of the trained text classifier machine learning models, a test set of input texts to form a plurality of sets of class determinations, wherein each set of class determinations corresponds to a respective one of the plurality of replacement method maps;

determine an accuracy of each of the plurality of sets of class determinations;

configure the domain terms replacer component with the replacement method map corresponding to a set of class determinations having a highest accuracy;

receive, by the data processing system, a training set of text and class specification pairs to be used as a ground truth for training a text classifier machine learning model for a text classifier;

identify, by the domain terms selector component executing on a processor of the data processing system, at least one domain term in the texts of the training set;

replace, by the domain terms replacer component executing on a processor of the data processing system, the at least one identified domain term in the texts of the training set with a corresponding replacement term to form a revised set of text and class specification pairs; and

train, by the text classifier trainer component executing on a processor of the data processing system, the text classifier machine learning model using the revised set to form a trained text classifier machine learning model,

wherein replacing the at least one identified domain term comprises, for each given identified domain term, determining a domain term type of the given identified domain term and replacing the given identified domain term based on selecting a replacement method associated with the domain term type by a replacement method map, wherein the trained text classifier machine learning model classifies an input text into one of a plurality of text classes selected from the group consisting of: sentiment, factoid, descriptive, and reasoned.

12. The apparatus of claim 11 , wherein for at least one identified domain term, the corresponding replacement term is an empty string.

13. The apparatus of claim 11 , wherein for at least one identified domain term, the corresponding replacement term is a different domain term substring comprising a description of what the identified domain term contains.

14. The apparatus of claim 13 , wherein the different domain term substring identifies an entity canonical name, an entity type, a relation, or a negative relation.

15. The apparatus of claim 11 , wherein the replacement method is identified from a set consisting of: replacing the given identified domain term with an empty string, replacing the given identified domain term with a different domain term substring, or not replacing the given identified domain term.