Feedback discriminator

Patent metadata

Field	Value
Publication number	US-12159233-B2
Application number	US-201916676632-A
Country	US
Kind code	B2
Filing date	Nov 7, 2019
Priority date	Nov 7, 2019
Publication date	Dec 3, 2024
Grant date	Dec 3, 2024

Abstract

Disclosed is a computer implemented method to identify relevant feedback. The method includes training, with a set of training data, a generative adversarial network (GAN), wherein the GAN includes a generator and a discriminator, and the set of training data is comprised of a plurality of feedback, each instance of feedback includes a value in a plurality of categories, the categories including a positivity value, and a validity value. The method also includes receiving, from a user, a new instance of feedback. The method further includes determining, by the discriminator, a first validity value and a first positivity value of the new instance of feedback. The method includes updating, in response to the first validity value being valid, and the first positivity value being negative, the GAN. The method further includes, displaying the new instance of feedback.

First claim

What is claimed is:

1. A computer-implemented method comprising:
training, with a set of training data, a generative adversarial network (GAN), wherein the GAN includes a generator and a discriminator, and the set of training data is comprised of a plurality of feedback, each instance of feedback includes a value in a plurality of categories, the categories including a positivity value, and a validity value, and the GAN is used by a chatbot to interact with a user;
receiving, in response to the training and by the chatbot, an input from the user;
determining, by the chatbot and in response to receiving the input from the user, an intent of the input;
generating, by the chatbot, a response to the input based on the intent;
receiving, from the user during an interaction with the chatbot and in response the receiving the input and the generating the response, a new instance of feedback, wherein the new instance of feedback is based on the intent and the response;
determining, by the discriminator, a first validity value and a first positivity value of the new instance of feedback, wherein the first validity value represents a likelihood the new instance of feedback is relevant to the input and the first positivity value represents a likelihood the chatbut misunderstood the input;
updating, in response to the first validity value being valid, and the first positivity value being negative, the GAN, wherein the first positivity value being negative indicates the chatbot did misunderstand the input; and
displaying the new instance of feedback.

2. The method of claim 1 , wherein updating the GAN comprises:
adding the new instance to a ground truth, wherein the ground truth includes the set of training data.

3. The method of claim 2 , further comprising:
receiving, from a second user, a second new instance of feedback;
determining, for the second new instance of feedback, by the discriminator, using the set of training data, the positivity value is negative, and the validity value is invalid; and
determining, in response to adding the new instance to the ground truth, for the second new instance of feedback, by the discriminator, using the ground truth, a second positivity value is negative, and a second validity value is valid.

4. The method of claim 1 , wherein the generator is trained by a first subset of the set of training data, the first subset comprising feedback from the set of training data where each validity value is invalid.

5. The method of claim 4 , wherein the discriminator is trained by a second subset of the set of training data, the second subset comprising feedback from the set of training data where each validity value is valid.

6. The method of claim 5 , wherein the discriminator is further trained by an output of the generator, wherein the output of the generator includes a fake feedback created by the generator.

7. The method of claim 1 , wherein the plurality of feedback is obtained from the user using a chatbot.

8. The method of claim 1 , further comprising:
deleting, a third instance of new feedback in response to the GAN determining the positivity value of the third instance of new feedback is positive.

9. The method of claim 1 , further comprising:
deleting, a fourth instance of new feedback in response to the GAN determining the validity value of the fourth instance of new feedback is irrelevant.

10. The method of claim 1 , wherein the method is performed by a feedback system, executing program instructions, and wherein the program instructions are downloaded from a remote data processing system.

11. A system comprising:
a processor; and
a computer-readable storage medium communicatively coupled to the processor and storing program instructions which, when executed by the processor, are configured to cause the processor to:
train, with a set of training data, a generative adversarial network (GAN), wherein the GAN includes a generator and a discriminator, and the set of training data is comprised of a plurality of feedback, each instance of feedback includes a value in a plurality of categories, the categories including a positivity value, and a validity value, and the GAN is used by a chatbot to interact with a user;
receive, in response to the training and by the chatbot, an input from the user;
determine, by the chatbot and in response to receiving the input from the user, an intent of the input;
generate, by the chatbot, a response to the input based on the intent;
receive, from the user during an interaction with the chatbot and in response the receiving the input and the generating of the response, a new instance of feedback, wherein the new instance of feedback is based on the intent and the response;
determine, by the GAN, a first validity value and a first positivity value of the new instance of feedback, wherein the first validity value represents a likelihood the new instance of feedback is relevant to the input and the first positivity value represents a likelihood the chatbot misunderstood the input;
update, in response to the first validity value being valid, and the first positivity value being negative, the GAN, wherein the first positivity value being negative indicates the chatbot did misunderstand the input; and
display the new instance of feedback.

12. The system of claim 11 , wherein updating the GAN comprises:
adding the new instance to a ground truth, wherein the ground truth includes the set of training data.

13. The system of claim 12 , wherein the program instructions are further configured to cause the processor to:
receive, from a second user, a second new instance of feedback;
determine, for the second new instance of feedback, by the discriminator, using the set of training data, the positivity value is negative, and the validity value is invalid; and
determine, in response to adding the new instance to the ground truth, for the second new instance of feedback, by the discriminator, using the ground truth, a second positivity value is negative, and a second validity value is valid.

14. The system of claim 11 , wherein the generator is trained by a first subset of the set of training data, the first subset comprising feedback from the set of training data where each validity value is invalid.

15. The system of claim 14 , wherein the discriminator is trained by a second subset of the set of training data, the second subset comprising feedback from the set of training data where each validity value is valid.

16. A computer program product, the computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by a processing unit to cause the processing unit to:
train, with a set of training data, a generative adversarial network (GAN), wherein the GAN includes a generator and a discriminator, and the set of training data is comprised of a plurality of feedback, each instance of feedback includes a value in a plurality of categories, the categories including a positivity value, and a validity value, and the GAN is used by a chatbot to interact with a user;
receive, in response to the training and by the chatbot, an input from the user;
receive, from a user, a new instance of feedback;
determine, by the chatbot and in response to receiving the input from the user, an intent of the input;
generate, by the chatbot, a response to the input based on the intent;
receive, from the user during an interaction with the chatbot and in response the receiving the input and the generating of the response, a new instance of feedback, w

Assignees

• IBM

Inventors

• Smith Aaron T
• James Rebecca Rose

Classifications

• G06N3/0475

  Generative networks · CPC title

• G06N3/094

  Adversarial learning · CPC title

• G06N3/09

  Supervised learning · CPC title

• G06N3/047

  Probabilistic or stochastic networks · CPC title

• G06N3/045

  Combinations of networks · CPC title