Adversarial training data augmentation for generating related responses
An intelligent computer platform to introduce adversarial training to natural language processing (NLP). An initial training set is modified with synthetic training data to create an adversarial training set. The modification includes use of natural language understanding (NLU) to parse the initial training set into components and identify component categories. As input is presented, a classifier evaluates the input and leverages the adversarial training set to identify the intent of the input. An identified classification model generates accurate and reflective response data based on the received input.
1. A computer system comprising:
a processing unit in communication with memory; and
an artificial intelligence (AI) platform in communication with the processing unit, the AI platform configured to support natural language processing, including:
a classifier, using natural language understanding (NLU), configured to evaluate received input, including:
process the received input against an adversarial training set;
leverage the adversarial training set, and predict a classification label of the received input;
use the predicted classification label to identify a corresponding classification model of the received input;
identify, by the classification model, an intent corresponding to the received input;
execute the identified classification model, including using the received input as characteristic data to the classification model; and
produce response data generated from the classification model execution, wherein the response data reflects the received input; and
a trainer operatively coupled to the classifier and configured to:
generate new synthetic training data using the received input;
narrow the new synthetic training data to a narrowed set, including to apply a log likelihood value to mathematically assess a proximity of the new synthetic training data to the intent corresponding to the received input; and
leverage the narrowed set of the new synthetic training data to dynamically augment training of the identified classification model.
2. The system of claim 1 , further comprising the classifier configured to assign the predicted classification label to the input and classify a semantic meaning of the received input.
3. The system of claim 1 , wherein the trainer is configured to modify the identified classification model with the received input and the generated response data.
4. The system of claim 3 , wherein the dynamically augmented classification model functions as a NLU domain classification index.
5. A computer program product comprising:
a computer readable storage medium having computer readable program code embodied therewith, the program code being executable by a processor to support natural language processing, including program code to:
use natural language understating (NLU) to process received input against an adversarial training set;
leverage the adversarial training set, and predict a classification label of the received input;
use the predicted classification label to identify a corresponding classification model of the received input;
identify an intent corresponding to the received to the received input;
execute the identified classification model, including using the received input as characteristic data to the classification model;
produce response data generated from the classification model execution, wherein the response data reflects the received input;
generate new synthetic training data using the received input;
narrow the new synthetic training data to a narrowed set, including to apply a log likelihood value to mathematically assess a proximity of the new synthetic training data to the intent corresponding to the received input; and
leverage the narrowed set of the new synthetic training data to dynamically augment training of the identified classification model.
6. The computer program product of claim 5 , further comprising program code executable by the processor to assign the predicted classification label to the input and classify a semantic meaning of the received input.
7. The computer program product of claim 5 , further comprising program code executable by the processor to modify the identified classification model with the received input and the generated response data.
8. The method of claim 7 , wherein the dynamically augmented classification model includes program code executable by the processor to function as a NLU domain classification index.
9. A method for processing natural language, comprising:
using natural language understanding (NLU), processing received input against an adversarial training set;
leveraging the adversarial training set, and predicting a classification label of the received input;
using the predicted classification label to identify a corresponding classification model of the received input;
identifying, by the classification model, an intent corresponding to the received input;
executing the identified classification model, including using the received input as characteristic data to the classification model;
producing response data generated from the classification model execution, wherein the response data reflects the received input;
generating new synthetic training data using the received input;
narrowing the new synthetic training data to a narrowed set, including applying a log likelihood value to mathematically assess a proximity of the new synthetic training data to the intent corresponding to the received input; and
leveraging the narrowed set of the new synthetic training data to dynamically augment training of the identified classification model.
10. The method of claim 9 , further comprising assigning the predicted classification label to the input and classifying a semantic meaning of the received input.
11. The method of claim 9 , further comprising modifying the identified classification model with the received input and the generated response data.
12. The method of claim 11 , wherein the dynamically augmented classification model functions as a NLU domain classification index.