Systems and methods for continual updating of response generation by an artificial intelligence chatbot
Methods and systems are provided for a natural language processing system comprising a chatbot adapted for dialog generation. In one example, the system may include a combination of a variational autoencoder (VAE) and a generative adversarial network (GAN) for generating natural responses to input queries. The VAE may convert queries into vector embeddings that may then be used by the GAN to continuously update and improve responses provided by the chatbot.
1. A method for training an artificial intelligence natural language processing system, comprising:
entering an input query into a user interface;
receiving an input query by a variational autoencoder (VAE);
embedding the input query into vector representations via the VAE;
converting the vector representations into responses by a generative adversarial network (GAN);
updating the GAN by a training process; and
displaying a response to the input query in a user interface based on the conversion provided by the GAN.
2. The method of claim 1 , further comprising creating, with an encoder of the VAE encoder, an embedding vector from the input query, and creating, with a decoder of the VAE, a response embedding from the embedding vector.
3. The method of claim 2 , wherein converting the vector representations into responses by the GAN includes sending the response embedding from the VAE to the GAN.
4. The method of claim 3 , further comprising receiving, at a generator of the GAN (GAN-G), the response embedding from the decoder, and generating, with the GAN-G, a response sentence to the input query.
5. The method of claim 4 , further comprising passing the response embedding received by the GAN-G from the decoder to a discriminator of the GAN (GAN-D), and receiving, at the GAN-D, data from a training data set including a seed answer to the input query.
6. The method of claim 5 , further comprising determining, with the GAN-D, if the received data originates from the GAN-D or the training data set, and calculating, with the GAN-D, losses based on a result of the determination.
7. The method of claim 6 , further comprising training the GAN-G by the GAN-D to generate more suitable responses based on the calculated losses.
8. The method of claim 7 , further comprising self-updating the GAN-D based on the calculated losses.
9. The method of claim 1 , wherein training the artificial intelligence natural language processing system includes updating the artificial intelligence natural language processing system based on question-answer pairs and user's responses.
10. The method of claim 1 , wherein training the artificial intelligence natural language processing system includes using a loss function to update the artificial intelligence natural language processing system.
11. A system for an artificial intelligence dialog generator, comprising;
a user interface device; and
a processor communicatively coupled to the user interface device, the processor configured with a variational autoencoder (VAE) combined with a generative adversarial network (GAN) stored in non-transitory memory, the processor further configured with instructions stored in the non-transitory memory that, when executed, cause the processor to:
accept input queries from a user interface;
calculate, using the VAE and GAN, responses to the queries; and
output the responses by displaying the responses in the user interface.
12. The system of claim 11 , wherein the VAE includes at least one neural network dense layer.
13. The system of claim 12 , wherein an encoder of the VAE is a recurrent neural network.
14. The system of claim 12 , wherein an encoder of the VAE is a convolutional neural network.
15. The system of claim 12 , wherein a decoder of the VAE is a convolutional neural network.
16. The system of claim 11 , wherein the GAN includes a generator that generates responses and a discriminator that discriminates between generated responses and anticipated responses.
17. A method for providing continually updating chatbot responses comprising;
training a dialog generator of the chatbot by a combination of a variation autoencoder (VAE) coupled to a generative adversarial network (GAN);
updating a discriminator of the GAN based on competition between a generator of the GAN and the discriminator of the GAN;
updating the VAE by backpropagation based on updating of the discriminator of the GAN; and
displaying a response provided by the updated VAE and the updated GAN.
18. The method of claim 17 , wherein training the dialog generator includes embedding an input query at the VAE as vectors, generating response embeddings from the vectors, and sending the response embeddings to the generator of the GAN to create a sentence based on the response embeddings, the created sentence evaluated against seeded answers at the discriminator of the GAN.
19. The method of claim 17 , wherein updating the discriminator of the GAN includes performing a loss calculation, using the loss calculation to adjust the discriminator to reduce the loss calculation, and updating the discriminator and the VAE through backpropagation.
20. The method of claim 17 , wherein displaying the response includes presenting an updated answer generated by the VAE and GAN in an output textbox shown at a display device.