Intention recognition method and intention recognition device

Abstract

The invention discloses an intention recognition method and an intention recognition device. The method comprises the following steps: acquiring a sentence to be recognized; converting the sentence tobe recognized into a first vector, a second vector and a third vector through an embedding layer; respectively performing linear mapping on the first vector and the second vector to obtain a first mapping vector and a second mapping vector; obtaining a first attention vector according to the first mapping vector, and obtaining a second attention vector according to the second mapping vector; obtaining a first sentence vector according to the first attention vector, and obtaining a second sentence vector according to the second attention vector; obtaining a target sentence vector according tothe first sentence vector and the second sentence vector; and inputting the target sentence vector into a linear layer of a Bert model for intention recognition, and obtaining an intention recognitionresult.

Description

technical field [0001] The present invention relates to the technical fields of natural language processing and neural network, in particular to an intention recognition method and an intention recognition device. Background technique [0002] The Transformer model is a natural language processing model (Natural Language Processing, referred to as NLP) developed by Google, and it is an important innovation in the development of NLP. The Bert model (BidirectionalEncoder Representations from Transformers) is based on the encoder structure of the Transformer model. The encoder includes multiple encoding layers, and one encoding layer includes a self-attention (self-attention) layer, addition & normalization layer and a fully connected feed-forward neural network layer, and the self-attention layer is formed by stacking N Scaled Dot Product Attention (SDPA) components, that is, SDPA is a component of the Bert model, and the Bert model uses an embedding layer to input sentences ...

AI Technical Summary

Problems solved by technology

[0007] Embodiments of the present invention provide an intent recognition method, an intent recognition device, and a computer-readable storage medium to solve the problem of limited performance of existing intent recognition and improve the accuracy of sentence intent recognition

Images