Liu, Jingxuan, Zhu, Ming, Ouyang, Huajiang 
ORCID: 0000-0003-0312-0326, Sun, Guozi and Li, Huakang
(2022)
Triplet Embedding Convolutional Recurrent Neural Network for Long Text Semantic Analysis.
In: Web Information Systems Engineering – WISE 2022.
|
Text
Liu WISE2022 TripletEmbeddingCNN.pdf - Author Accepted Manuscript Download (861kB) | Preview |
Abstract
Deep Recurrent Neural Network has an excellent performance in sentence semantic analysis. However, due to the curse of the computational dimensionality, the application in the long text is minimal. Therefore, we propose a Triplet Embedding Convolutional Recurrent Neural Network for long text analysis. Firstly, a triplet from each sentence of the long text. Then the most crucial head entity into the CRNN network, composed of CNN and Bi-GRU networks. Both relation and tail entities are input to a CNN network through three splicing layers. Finally, the output results into the global pooling layer to get the final results. Entity fusion and entity replacement are also used to retain the text’s structural and semantic information before triplet extraction in sentences. We have conducted experiments on a large-scale criminal case dataset. The results show our model significantly improves the judgment prediction task.
| Item Type: | Conference or Workshop Item (Unspecified) |
|---|---|
| Uncontrolled Keywords: | Mental Health |
| Depositing User: | Symplectic Admin |
| Date Deposited: | 19 Dec 2022 10:02 |
| Last Modified: | 15 Mar 2024 05:42 |
| DOI: | 10.1007/978-3-031-20891-1_43 |
| Related URLs: | |
| URI: | https://livrepository.liverpool.ac.uk/id/eprint/3166711 |
Dimensions
Dimensions
