Elegant Sentences Recognition Based on Part of Speech Features and Hybrid Neural Network
WANG Zhiheng
WANG Fang
HUANG Shucheng
Abstract：This paper proposes a recognition model for Chinese composition graceful sentences based on part-of-speech fea-tures and hybrid neural networks.Firstly,the terms in the sentence and the part-of-speech features corresponding to each term are vectorized and represented by fusion.Secondly,the convolutional neural network(CNN)with the introduction of attention mecha-nisms is used to obtain the local features of the fused word vectors.Bi-Gated recurrent unit(BiGRU)is used to serialize the fused word vectors to obtain global features.This paper fuses the two features and connects them to a classifier for graceful sentence recog-nition.Experiments show that the model based on part-of-speech features and hybrid neural network has the highest recognition ac-curacy for beautiful sentences,reaching 90.07%and F1 value reaching 83.13%.
Keywords：elegant sentencespart-of-speech featuresCNNattention mechanismsBiGRU
Publication Date：2025-04-20
Online Publishing Date：2025-08-15(First online date of this platform, not the publication date of the document)
Pages：7( 995-1001 )