Task-oriented dialogue strategy generation method

Abstract

The invention relates to a task-oriented dialogue strategy generation method, and the method comprises the following steps: establishing a dialogue state tracker, and determining a dialogue state space, an action space and formalized representation thereof; simulating a dialogue state by using a variational automatic encoder; simulating a dialogue action by using a multi-layer perceptron and Gumbel Softmax; performing adversarial training on a simulation sample generator and a discriminator; and finally training a dialogue strategy by using a reinforcement learning method. Firstly, a simulation sample generator is used for learning a reward function, and loss from a discriminator can be directly fed back to the generator for optimization; secondly, the trained discriminator is taken as a dialogue reward to be brought into a reinforcement learning process for guiding dialogue strategy learning; the dialogue strategy can be updated by utilizing any reinforcement learning algorithm; according to the method, common information contained in high-quality dialogues generated by human beings can be deduced by distinguishing the dialogues generated by the human beings and the machine respectively, and then the learned information is fully utilized to guide dialogue strategy learning in a new field in a transfer learning mode.

Description

technical field [0001] The invention relates to a method for generating a task-oriented dialogue strategy, which belongs to the technical field of natural language processing. Background technique [0002] A task-oriented dialogue system is designed to provide users with services to complete specific tasks, such as booking a hotel, buying a movie ticket, etc.; this dialogue system requires a specific dialogue strategy in order to select the most appropriate dialogue strategy in each dialogue round according to the context of the current dialogue. Actions. [0003] The development of reinforcement learning in robotics and other fields has brought new inspiration to dialogue strategy learning. After clarifying the state space and action space, the goal of task-oriented dialogue systems is to maximize positive feedback from users. The dialogue policy learning method based on reinforcement learning is suitable for training with user simulators instead of real people in order to...

AI Technical Summary

Problems solved by technology

But this approach is limited to policy gradient-based algorithms that alternately update dialogue policy and reward models; non-policy gradient methods cannot benefit from self-learned reward functions

In addition, the alternating cycle of dialogue strategy and reward model can easily fall into a local optimal state or lead to model collapse.

Images