Multi-round dialogue processing method for professional scenes

Abstract

The invention relates to the technical field of human-computer interaction, and aims to provide a multi-round dialogue processing method for professional scenes. The method comprises the following steps: constructing a knowledge graph, wherein the knowledge graph comprises a first-level node, a second-level node, a third-level node and a fourth-level node; generating a semantic slot rule for multiple rounds of conversations based on the knowledge graph content corresponding to the scene information entity; receiving current problem information, and identifying a scene information entity of a current multi-round dialogue in the current problem information; and performing subgraph search operation in the knowledge graph based on the first-level node, the second-level node, the third-level node and the fourth-level node in the scene information entity of the current multi-round dialogue to obtain the knowledge graph corresponding to the current scene information entity, and then outputting guidance questions and answers or outputting final response replies based on semantic slot rules. The problem of structural database data redundancy is solved, the effective information retrieval speed is increased, meanwhile, the problem that a single scene corresponds to a single multi-round dialogue is solved, and the labor cost is saved.

Description

technical field [0001] The invention relates to the technical field of human-computer interaction, in particular to a multi-round dialogue processing method for professional scenes. Background technique [0002] Existing multi-round dialog processing methods mainly include methods based on rules, retrieval and production. In the process of using the prior art, the inventors have found that at least the following problems exist in each prior art: a. Method based on rules: In rule-based In multiple rounds of conversations, it is necessary to define a set of rule models in advance (such as logical judgment, keyword or word search, or some more complex classifiers), and the rule model is based on the set rules and the information extracted from the previous conversation information, evaluate the input question, and perform corresponding operations. The operation of this mode is relatively simple, but for different data sources, the required rule models are often not the same, s...

AI Technical Summary

Problems solved by technology

This method does not need to define the corresponding reply text set in advance, and directly generates the reply text, but requires a large amount of training data, the model training cost is relatively high, and the generated reply may be far from the user's expected answer (there are certain grammatical errors or the reply deviates from the question itself), and there is also the problem of inconsistency in the reply

Images