Conversation strategy generation method for simulating user emotion

Abstract

The invention discloses a dialogue strategy generation method for simulating user emotion, and the method comprises the following steps: establishing a user character and emotion representation system; representing an environment triggering event; establishing function representation of emotion changes in the dialogue; and finally, selecting a dialogue action in combination with the user emotion. By simulating more complex emotional states and change processes for people with different personalities, the generalization ability of a dialogue strategy model is improved, and a dialogue system is allowed to formulate targeted dialogue strategies for users with different personalities; the dialogue strategy generation method can simulate that the user terminates the dialogue at any time in the dialogue instead of always assuming the user as an ideal individual which always cooperates with the dialogue, so that a simulation behavior which is closer to an actual user than an existing dialogue system user simulator can be generated, thereby allowing the training and evaluation of dialog policies to the robot in a simulation framework that is closer to the real environment.

Description

technical field [0001] The invention relates to a dialogue strategy generation method for simulating user emotions. Background technique [0002] According to the number of rounds of interaction, the human-computer dialogue system can be divided into two forms: single round and multi round. Single-round dialogue is often a traditional question-and-answer system. The question includes the necessary information needed to find the answer, and the system gives an answer based on the information in the question. The form of multi-round dialogue is closer to the interaction habits of real people, but at the same time it is more difficult to realize. A typical multi-round dialogue system includes three main components: natural language understanding, dialogue management, and natural language generation. The dialogue management component includes two parts: dialogue state tracking and dialogue strategy. Understand and collect all kinds of information. These information sequences c...

AI Technical Summary

Problems solved by technology

The main reason is that deep learning-based methods usually require a large amount of data for model learning, but the existing publicly labeled task-oriented dialogue data usually does not contain emotional information; at the same time, the existing user simulators also ignore the simulation of user emotions, Therefore, it is difficult to effectively study task-oriented dialogue policy generation methods involving user emotions.

Images