TY - GEN
T1 - Triplet Embedding Convolutional Recurrent Neural Network for Long Text Semantic Analysis
AU - Liu, Jingxuan
AU - Zhu, Ming
AU - Ouyang, Huajiang
AU - Sun, Guozi
AU - Li, Huakang
N1 - Publisher Copyright:
© 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.
PY - 2022
Y1 - 2022
N2 - 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.
AB - 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.
KW - CRNN
KW - Entity fusion and entity replacement
KW - Long text analysis
KW - Triplet embedding
UR - http://www.scopus.com/inward/record.url?scp=85142681247&partnerID=8YFLogxK
U2 - 10.1007/978-3-031-20891-1_43
DO - 10.1007/978-3-031-20891-1_43
M3 - Conference Proceeding
AN - SCOPUS:85142681247
SN - 9783031208904
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 607
EP - 615
BT - Web Information Systems Engineering – WISE 2022 - 23rd International Conference, Proceedings
A2 - Chbeir, Richard
A2 - Huang, Helen
A2 - Silvestri, Fabrizio
A2 - Manolopoulos, Yannis
A2 - Zhang, Yanchun
A2 - Zhang, Yanchun
PB - Springer Science and Business Media Deutschland GmbH
T2 - 23rd International Conference on Web Information Systems Engineering, WISE 2021
Y2 - 1 November 2022 through 3 November 2022
ER -
