Frame semantic parsing is a fundamental task in natural language processing. Most of the previous research work focuses on the design of the single-target model. Therefore, these models can't extract frame semantic structures of multiple targets in one time. This paper designs a frame semantic parsing model for multiple targets which jointly predicts the results of multiple targets. We model and achieve the bidirectional interaction among the sub tasks of frame semantic parsing. Moreover, Relational Graph Convolution Network (R-GCN) is utilized to encode the frame relation information, which is a way to exploit frame semantic knowledge into our model. The experiments shows that our model maintains good performance without extra training corpus. Ablation Study proves the effectiveness of our model.
框架语义分析；框架网络
陈旭东、郑策、常宝宝
展开 >
北京大学计算语言学教育部重点实验室，北京100871,北京大学软件与微电子学院，北京102600
北京大学计算语言学教育部重点实验室，北京100871
框架语义分析；框架网络
Chinese national conference on computational linguistic
Nanchang(CN)
The 21st Chinese national conference on computational linguistic
NETL
