Chinese text intention recognition method based on Bert and full-connection neural network fusion

Abstract

The invention discloses a Chinese text intention recognition method based on Bert and full-connection neural network fusion, and the method comprises the steps: carrying out the text preprocessing of a plurality of Chinese intention corpora in a Chinese intention corpus data set, and carrying out the word segmentation with a character as a unit; finding a vector expression of a corresponding word from the word vector table and converting each word into a vector with a fixed length; respectively extracting sentence embedding characteristics and position embedding characteristics of each intention text by utilizing a Bert pre-training model; employing a Bert bidirectional transformer encoder for extracting features to obtain a sequence vector containing semantic features, and taking a full-connection neural network model as an identification model to realize intention identification. According to the invention, a Bert pre-training model is used for enabling a machine to have priori knowledge; therefore, the method can be used for solving the problem that the recognition precision is reduced due to relatively single semantic features in text intention recognition.

Description

technical field [0001] The invention belongs to the technical field of natural language processing, in particular to a Chinese text intent recognition method based on the fusion of Bert and a fully connected neural network. Background technique [0002] With the rapid development of the Internet and artificial intelligence, it has gradually become a reality for humans to be replaced by machines, among which dialogue commerce is developing rapidly. People like to express their feelings and needs in natural language. Conversational commerce technology enables enterprises to chat with customers on the platform that customers prefer, optimizing expressiveness, relevance and personalization. Under the wave that human-computer interaction has shifted from graphic window interaction to dialogue window interaction, natural language processing technology has become a key element of human-computer interaction. However, it is still an urgent challenge for machines to communicate with ...

AI Technical Summary

Problems solved by technology

However, it is still an urgent challenge for machines to communicate with humans to complete specific tasks. One of the difficulties to be overcome is to allow machines to recognize the user's conversational intent

[0003] Traditional intent recognition methods use non-dynamic word vectors or word vectors as semantic features, and the amount of contextual information is relatively single. Most of the feature extraction models used are CNN and RNN models in deep learning, which are not capable of extracting corpus features.

Images