Task-oriented dialog system and method through feedback
AssigneeBeijing Didi Infinity Technology and Development Co., Ltd.
InventorScot FANG
Life of this patent
Estimating the term…
Abstract
An automatic agent may be improved through feedback. A user input may be received through a user interface. A plurality of current utterance variables may be obtained by tokenizing the user input. The automatic agent may execute a machine learning policy to generate a reply to the user input based on the plurality of current utterance variables. A different reply may be obtained in response to an indication that the reply will lead to a breakdown, wherein the breakdown comprises an unhuman response from the automatic agent according to the machine learning policy. The machine learning policy may be adjusted based on the plurality of current utterance variables and the different reply.
Claims20
A computer-implemented method for improving an automatic agent, comprising:
- receiving a user input through a user interface;
- obtaining a plurality of current utterance variables by tokenizing the user input;
- generating a reply to the user input based on the plurality of current utterance variables and the automatic agent executing a machine learning policy;
- obtaining a different reply in response to an indication that the reply will lead to a breakdown, wherein the breakdown comprises an unhuman response from the automatic agent according to the machine learning policy;
- and
- adjusting the machine learning policy based on the plurality of current utterance variables and the different reply.
A system for improving an automatic agent, comprising one or more processors and one or more non-transitory computer-readable memories coupled to the one or more processors and configured with instructions executable by the one or more processors to cause the system to perform operations comprising:
- receiving a user input through a user interface;
- obtaining a plurality of current utterance variables by tokenizing the user input;
- generating a reply to the user input based on the plurality of current utterance variables and the automatic agent executing a machine learning policy;
- obtaining a different reply in response to an indication that the reply will lead to a breakdown, wherein the breakdown comprises an unhuman response from the automatic agent according to the machine learning policy;
- and
- adjusting the machine learning policy based on the plurality of current utterance variables and the different reply.
A non-transitory computer-readable storage medium configured with instructions executable by one or more processors to cause the one or more processors to perform operations comprising:
- training a machine learning policy based on a training corpus, wherein the training corpus comprises a plurality of dialog samples;
- generating a reply to a user input based on an automatic agent executing the machine learning policy;
- adjusting the machine learning policy based on a different reply in response to an indication that the reply will lead to a breakdown;
- and
- deploying the automatic agent to interact with human users based on the adjusted machine learning policy.
Dependencies are inferred from the claim text.