Dialogue generation system and dialogue implementation method

Abstract

The invention discloses a dialogue generation system and a dialogue implementation method. According to the dialogue generation system provided by the embodiment of the invention, a neural network ofan encoder-decoder structure is adopted, double coding layers for a character level and a word level respectively are adopted in a coding layer, and information coded by the double coding layers is combined. Before a data set containing a dialogue target, related knowledge information and a dialogue sequence is encoded by an encoding layer, data sets for a character-level coding layer and a word-level coding layer are reconstructed respectively, the reconstructed data sets are input into double coding layers in a dialogue generation system for coding and combination respectively, the coded andcombined data sets are input into a decoding layer, and corresponding dialogue replies are generated by the decoding layer according to combined data set information by using a transformer model. Therefore, the accuracy and fluency of the generated corresponding dialogue reply are improved.

Description

technical field [0001] The invention relates to the field of computer technology, in particular to a dialog generation system and a dialog realization method. Background technique [0002] Human-machine dialogue technology is still in its infancy. The entire interaction process is mostly led by the user, and the machine responds passively, that is, the machine's reply is used to respond to the user's input, and it cannot conduct a self-directed dialogue interaction like a human. In order to make the dialogue and interaction dominated by the machine, an active chat task based on the knowledge graph is set up. The machine conducts active chatting according to the constructed knowledge graph, so that the machine has the ability to simulate human beings to use language to transmit information. [0003] During the conversation, the machine is active, guiding the user from one topic to another. The dialogue system sets a dialogue goal for the machine. The dialogue goal is "START→...

AI Technical Summary

Problems solved by technology

[0006] 1) The encoding layer is independent of each other when encoding, ignoring the possibility that the historical dialogue sequence can be integrated to make the semantics of the dialogue sequence richer;

[0007] 2) The decoding layer lacks in-depth interaction when determining the corresponding dialogue reply, and at the same time appears too thin, which may lead to insufficient reasoning or information loss

[0008] To sum up, when using the current dialogue system to process dialogues, there are problems that the corresponding dialogue responses obtained are not accurate and smooth

Images