Emotional dialog generation system and method

Abstract

The invention relates to an emotional dialogue generation system and method. The emotional dialogue generation system comprises an emotional dialogue generation module and a reordering module; the emotional dialogue generation module comprises a basic reply generation module, a multi-model emotion reply generation module and a single-model emotion reply generation module; the basic reply generation module is used for generating a basic reply with correct semantics; the multi-model emotion reply generation module is used for establishing emotion models corresponding to various emotions throughhierarchical training and obtaining emotion replies based on the emotion models; the single-model emotion reply generation module takes the emotion category as input to train the emotion models and outputs an emotion reply according to the emotion models; the reordering module receives the replies output by three sub-modules in the dialogue generation module, scores the output replies, and rearranges the replies output by the modules according to the scores, and the reply with the highest score is the final emotion reply. In the man-machine dialogue process, the machine can generate dialogue replies meeting specific emotions, grammars are smooth, semantics are coherent, emotion consistency is met, and thus the user experience feeling in man-machine interaction is improved.

Description

technical field [0001] The invention relates to an emotional dialogue generation system and method, and belongs to the technical field of artificial intelligence. Background technique [0002] With the emergence of human-computer interaction products such as Siri, Google Home, Tmall Genie, Xiao Ai and other personal assistants, smart speakers, etc., human-computer interaction has not only attracted more and more attention from the industry and academia, but also has an impact on people's daily life. growing impact. At present, there are many ways to realize the dialogue system, but the dialogue mode of the dialogue system is mostly task-based dialogue, that is, the human-computer interaction is carried out by issuing commands. However, with the development of science and technology, simple task-based dialogue is no longer what people need. People hope that the reply content of the robot can be more fluent, in line with the way people speak, and be able to distinguish the em...

AI Technical Summary

Problems solved by technology

Retrieval-based or template-based dialogue models can avoid irrelevant answers and grammatical errors, but their practical application is often limited by the quality of the matching model and the system cannot contain all reply sentences or templates

As a result, the machine's replies are usually relatively blunt, rigid, not rich in semantics, and have no emotional color

Especially for some relatively new vocabulary, expressions and some obvious emotional colors, it is impossible to give semantic answers

Images