System, method, and computer program for real-time language translation using generative artificial intelligence
As described herein, a system, method, and computer program provide real-time language translation using generative artificial intelligence. An input in a first spoken language is received. The input in the first spoken language is processed, using a generative artificial intelligence model, to generate a translation of the input in a second spoken language. The translation is output.
1 . A non-transitory computer-readable media storing computer instructions which when executed by one or more processors of a device cause the device to:
receive, via an application programming interface (API) enabling communications between an application and a language translation service, an input that includes:
audio of a speech that has been generated in a first spoken language by a chatbot process of the application, and
an indication of a second spoken language to which the speech is to be translated;
process the input, using a generative artificial intelligence model of the language translation service, to generate in real time a translation of the speech in the second spoken language; and
output, by the language translation service to the application, an audio of the translation in the second spoken language to cause the chatbot process to operate as a multilingual chatbot that communicates in real time with a customer in the second spoken language.
2 . The non-transitory computer-readable media of claim 1 , wherein the device is further caused to:
generate a training data set for use in training the generative artificial intelligence model.
3 . The non-transitory computer-readable media of claim 2 , wherein the training data set includes multilingual text and speech data in a standardized format.
4 . The non-transitory computer-readable media of claim 3 , wherein the multilingual text and speech data is collected from at least one of:
public databases,
online corpora, or
user generated content.
5 . The non-transitory computer-readable media of claim 2 , wherein the device is further caused to:
train the generative artificial intelligence model using the training data set.
6 . The non-transitory computer-readable media of claim 1 , wherein the generative artificial intelligence model is a transformer model.
7 . The non-transitory computer-readable media of claim 6 , wherein the generative artificial intelligence model is a Long Short-Term Memory network (LSTM).
8 . The non-transitory computer-readable media of claim 1 , wherein the generative artificial intelligence model is a Convolutional Neural Network (CNN).
9 . The non-transitory computer-readable media of claim 1 , wherein the generative artificial intelligence model is an Encoder-Decoder model.
10 . The non-transitory computer-readable media of claim 1 , wherein the generative artificial intelligence model is a neural machine translation (NMT) model.
11 . The non-transitory computer-readable media of claim 1 , wherein the receiving, processing, and outputting is performed in real-time.
12 . The non-transitory computer-readable media of claim 1 , wherein the device is further caused to:
test the translation with real-world data, and
update the generative artificial intelligence model based on a result of the testing.
13 . The non-transitory computer-readable media of claim 12 , wherein the real-world data includes feedback from a user regarding the translation.
14 . The non-transitory computer-readable media of claim 13 , wherein the generative artificial intelligence model is updated using reinforcement learning.
15 . A method, comprising:
at a computer system:
receiving, via an application programming interface (API) enabling communications between an application and a language translation service, an input that includes:
audio of a speech that has been generated in a first spoken language by a chatbot process of the application, and
an indication of a second spoken language to which the speech is to be translated;
processing the input, using a generative artificial intelligence model of the language translation service, to generate in real time a translation of the speech in the second spoken language; and
outputting, by the language translation service to the application, an audio of the translation in the second spoken language to cause the chatbot process to operate as a multilingual chatbot that communicates in real time with a customer in the second spoken language.
16 . A system, comprising:
a non-transitory memory storing instructions; and
one or more processors in communication with the non-transitory memory that execute the instructions to:
receive, via an application programming interface (API) enabling communications between an application and a language translation service, an input that includes:
audio of a speech that has been generated in a first spoken language by a chatbot process of the application, and
an indication of a second spoken language to which the speech is to be translated;
process the input, using a generative artificial intelligence model of the language translation service, to generate in real time a translation of the speech in the second spoken language; and
output, by the language translation service to the application, an audio of the translation in the second spoken language to cause the chatbot process to operate as a multilingual chatbot that communicates in real time with a customer in the second spoken language.