A sensitive multi-round dialogue management system and method based on state machine context

Abstract

A sensitive multi-round dialogue management system based on state machine contexts comprises the following units: an input module used for receiving multi-modal input information of a user; an intention identification engine module for identifying intent information in the multi-modal input information; an intention module mapping the identified plurality of intent information with the back-end multiple intension sub-modules one by one; a state machine module comprising a plurality of state machines and used for managing related contexts in a dialogue manage system, and supporting an output result; an analysis instruction engine module including a plurality of analysis instruction engine sub-modules and used for respectively analyzing the corresponding intention information and obtaining the plurality of intention information after the analysis; an output module output module obtaining the policy information according to the result of the analysis of the intention identification module and the analysis engine module, and sending the policy information to the state machine module. The dialogue management system improves the probability of success of man-machine interaction, and can ensure that man-machine conversations are effective and friendly carried out.

Description

technical field [0001] The invention relates to a dialogue management system, in particular to a context-sensitive multi-round dialogue management system and method based on a state machine. Background technique [0002] Multi-round interaction is the most important part of the input dialogue system, and it is not only applicable to the input dialogue system, but also applicable to all dialogue management mode scenarios. The existing dialog management methods mainly include: most of them use rule-based methods to construct, such as slot filling method, finite automata method, etc. This kind of rule-based human-computer dialog model has been successfully applied in business. [0003] Dialogue management technologies based on statistical models include: Bayesian networks, graphical models, dialogue-based reinforcement learning techniques, partially observable Markov decision process (POMDP), etc., which enable computers to flexibly handle human-machine dialogues User input er...

AI Technical Summary

Problems solved by technology

Although this method is novel, the uncertainty of the user model is great, the automaton transition conditions described are too complicated, and the state definition is not clear

Images