Conversation generation method and device

Abstract

The embodiment of the invention discloses a conversation generation method, which comprises the following steps that: converting each character in the Kth round of inquiry sentence into a first word vector, and calculating the forward direction implicit strata vector and the reverse direction implicit strata vector of each character according to the first word vector; obtaining the content topic of the Kth round of inquiry sentence, and converting the content topic into a second word vector; according to the second word vector, the forward direction implicit strata vector of the last characterin the Kth round of inquiry sentence, the implicit strata vector of the last character in a (K-1)th round of inquiry sentence output by aiming at a (K-1)th round of inquiry sentence and the initial implicit strata vector of the (K-1)th round of inquiry sentence output by aiming at a (K-1)th round of inquiry sentence, determining the initial implicit strata vector which aims at the Kth round of inquiry sentence; and according to the forward direction implicit strata vector and the reverse direction implicit strata vector of each character in the Kth round of inquiry sentence and the initial implicit strata vector which is output by aiming at the Kth round of inquiry sentence, generating a replying sentence which aims at the Kth round of inquiry sentence. When the embodiment of the invention is implemented, accuracy for generating dialogue can be improved.

Description

technical field [0001] The present invention relates to the field of voice processing, in particular to a dialog generation method and device. Background technique [0002] In recent years, the way of human-computer interaction is changing rapidly. As a new interaction mode, dialogue robot is attracting more and more people's attention. However, how to improve the relevance of automatically generated reply sentences in multiple rounds of dialogue, how to reduce the generation probability of high-frequency answers, and generate high-quality dialogues have always been the focus of research in the field of natural language processing. Among them, the dialogue system is a natural language An important application direction of processing. [0003] In the prior art solution, the dialogue system may include a rule-based dialogue system, a search-based dialogue system or a generative dialogue system. Among them, the rule-based dialogue system has a simple structure and high accura...

AI Technical Summary

Problems solved by technology

It can solve the technical problem of low accuracy in generating dialogs

Images