A semantic and syntactic enhanced neural model for financial sentiment analysis
作者:
Highlights:
• We show it is more challenging for the FSA than TBSA.
• SSENM employs semantic and syntactic knowledge to capture discriminative features.
• E-GCN enables each node representation to realize in-depth dependency connections.
• We integrate the model-based and data-driven approach to construct the regressor.
• Quality assessment methods are compared in two publicly available datasets.
摘要
•We show it is more challenging for the FSA than TBSA.•SSENM employs semantic and syntactic knowledge to capture discriminative features.•E-GCN enables each node representation to realize in-depth dependency connections.•We integrate the model-based and data-driven approach to construct the regressor.•Quality assessment methods are compared in two publicly available datasets.
论文关键词:Financial sentiment analysis,Attention mechanism,Graph convolutional network,Manifold mixup
论文评审过程:Received 15 November 2021, Revised 6 April 2022, Accepted 9 April 2022, Available online 10 May 2022, Version of Record 10 May 2022.
论文官网地址:https://doi.org/10.1016/j.ipm.2022.102943