Domain prediction method, domain prediction device and electronic equipment

Abstract

The invention provides a domain prediction method, a domain prediction device and electronic equipment. The domain prediction method comprises the following steps: determining a current round of interactive text; inputting the interactive text and the supervision information of this round into a domain prediction model; wherein the supervision information is obtained by correcting domain probability distribution which is output by the domain prediction model and corresponds to the last round of interactive text on the basis of domain information determined after semantic comprehension of the last round of interactive text, and the domain probability distribution is output by the domain prediction model and corresponds to the last round of interactive text; and determining a domain prediction result based on the domain probability distribution corresponding to the current round of interactive text. According to the domain prediction method provided by the embodiment of the invention, the accuracy of model prediction in the multi-round interaction process can be greatly improved, and particularly for simplified interaction in the multi-round interaction process, an accurate domain prediction result can be obtained.

Description

technical field [0001] The present invention relates to the technical field of voice interaction, and more specifically, to a field prediction method, a field prediction device and electronic equipment. Background technique [0002] In the process of voice interaction, in order to better understand semantics, it is usually necessary to predict which field the user's expression belongs to. [0003] In the prior art, one method is to match whether the user's expression belongs to a certain field based on the grammar rule network or state machine rules. This method has poor generalization ability, and it will not be able to understand the sentence patterns that are not included. Another way is to use the deep neural network to learn the sentence information of the user's expression content, so as to achieve the purpose of the model prediction field. However, the accuracy of the deep neural network in the prediction of the expression content of a single round of interaction is a...

AI Technical Summary

Problems solved by technology

[0003] In the prior art, one method is to match whether the user's expression belongs to a certain field based on the grammar rule network or state machine rules. This method has poor generalization ability, and it will not be able to understand the sentence patterns that are not included.

Another way is to use the deep neural network to learn the sentence information of the user's expression content, so as to achieve the purpose of the model prediction field. However, the accuracy of the deep neural network in the prediction of the expression content of a single round of interaction is acceptable. Once it involves After multiple rounds of dialogue, the accuracy of model predictions is greatly reduced

Images