Back to results
US 2020/0027443 A1Published application
Task-independent conversational systems
Assigneebotbotbotbot Inc.
InventorAntoine Raux
Life of this patent
Estimating the term…
Abstract
Methods, systems, and apparatus, including computer programs encoded on computer storage media, for generating responses using task-independent conversational systems.
Claims21
A method comprising:
- obtaining, by a computer-implemented dialogue system and during a conversation between the computer-implemented dialogue system and a user to complete a particular task, a text input from the user;
- obtaining, by the computer-implemented dialogue system, a state of the conversation with the user that has been updated based on the text input;
- generating, by the computer-implemented dialogue system and from the state, a conversational output that defines a conversational response to the text input that is independent of the particular task;
- and
- determining, by the computer-implemented dialogue system and from at least one of the conversational output or the state of the conversation, whether to (i) provide the conversational response to the text input or (ii) generate a task output that is specific to the particular task.
2 more elements
A method comprising:
- obtaining, by a computer-implemented dialogue system and during a conversation between the computer-implemented dialogue system and the user to complete a particular task, a text input from the user;
- obtaining, by the computer-implemented dialogue system, a state of the conversation with the user that has been updated based on the text input;
- generating, by the computer-implemented dialogue system and from the state, a conversational output that defines a conversational response to the text input that is independent of the particular task;
- generating, by the computer-implemented dialogue system and from the state, a task output that defines a task response to the text input that is specific to the particular task;
- generating, by the computer-implemented dialogue system and based on the conversational output and the task output, a final response to the text input;
- and
- providing, by the computer-implemented dialogue system, the final response to the user.
4 more elements
A method comprising:
- obtaining, by a computer-implemented dialogue system and during a conversation between the computer-implemented dialogue system and the user to complete a particular task, a text input from the user;
- obtaining, by the computer-implemented dialogue system, a state of the conversation with the user that has been updated based on the text input;
- generating, by the computer-implemented dialogue system and from the state, a task output that defines a task response to the text input that is specific to the particular task;
- determining, by the computer-implemented dialogue system and from the task output, that responding to the text input requires a conversational output that is shared between multiple tasks;
- in response, generating, by the computer-implemented dialogue system and from the state, a conversational output that defines a conversational response to the text input that is independent of the particular task;
- and
- generating a final response to the text input from the task output and the conversational output.
4 more elements
A method comprising:
- obtaining multi-task training data, the multi-task training data comprising a plurality of sequences of conversational inputs, wherein each sequence corresponds to a respective task, and the multi-task training data comprises sequences corresponding to multiple different tasks;
- and
- training a conversational machine learning model on the multi-task training data to determine trained values of the parameters of the conversational machine learning model, wherein the conversational machine learning model is configured to receive as input a conversational input and to generate as output a conversational output that defines a response to a user that is independent of a task being performed when the conversational input was generated.
Dependencies are inferred from the claim text.