Named entity recognition method and device based on hybrid lattice self-attention network

Abstract

The invention discloses a named entity recognition method based on a mixed lattice self-attention network, and the method comprises the steps: S1, encoding a sentence feature vector represented by a word pair into a matrix with a fixed dimension, and obtaining a word vector representation of a mixed lattice structure; constructing a self-attention network to capture the influence of a word vector on a word vector in the vector, and enhancing the feature representation of each word vector; word features are fused in an Embedding layer of BERT, and better word vector representation is obtained through learning by fine tuning of a learning process; an entity sequence labeling task and a decoding process in entity recognition are realized according to a BiLSTM-CRF network, modeling of fused character features is completed through the network, and an entity recognition model based on the hybrid lattice self-attention network is constructed. According to the method, global vocabulary information can be captured, semantic-rich word vector representation is generated, and the Chinese named entity recognition precision is improved on multiple data sets.

Description

technical field [0001] The invention relates to the technical field of natural language processing in artificial intelligence, in particular to a named entity recognition method and device based on a mixed lattice self-attention network. Background technique [0002] Named Entity Recognition (NER) is also called entity extraction. It was first proposed at the MUC-6 conference. It is a technology for extracting entities from text in information extraction technology. Early entity recognition used rule-based and statistics-based methods. Because these traditional methods rely too much on manual design, and the recognition coverage is small and the recognition accuracy is low, they have long been replaced by deep learning methods. In the method based on deep learning, the entity recognition model is divided into a character-based model (character-based) and a word-based model (word-based), and some other languages ​​such as English usually use a character-based model, because e...

AI Technical Summary

Problems solved by technology

[0004] Although the existing methods have achieved good results in the fusion of word feature vectors, there are problems in the existing technical means: 1) The word feature fusion method does not take into account the semantic expression of word vectors trained by different models. 2) In the vocabulary enhancement method based on learning word weights, only the matching word pairs of each word feature are considered. The impact of semantic representation, ignoring the role of global lexical information

Images