基于GCN的异构句法驱动汉语语义角色标注  

HETEROGENEOUS SYNTAX-AWARE SEMANTIC ROLE LABELING BASED ON GRAPH CONVOLUTIONAL NETWORKS

作　　者：杨浩苹 夏庆荣 李正华[1] 王睿 Yang Haoping;Xia Qingrong;Li Zhenghua;Wang Rui(School of Computer Science and Technology,Soochow University,Suzhou 215006,Jiangsu,China;Vipshop,Guangzhou 510000,Guangdong,China)

机构地区：[1]苏州大学计算机科学与技术学院,江苏苏州215006 [2]唯品会,广东广州510000

出　　处：《计算机应用与软件》2025年第4期122-128,134,共8页Computer Applications and Software

基　　金：国家自然科学基金项目(61876116)。

摘　　要：近年来,句法驱动的语义角色标注(Semantic Role Labeling,SRL)受到了广泛的关注。但是,大部分工作只考虑如何利用单个同构句法树信息。考虑到汉语中存在多个人工构建的高质量异构句法树库,提出采用图卷积神经网络(Graph Convolutional Networks,GCN)来刻画多个异构句法树中包含的句法信息,并深入比较多种编码方式,来提升汉语SRL的性能。最终,该模型在CPB 1.0和CoNLL-2009汉语数据集上分别达到了84.16%和85.30%的F1值,均高于编码同构句法树的实验结果,且相比于前人的方法有了显著的提升。Recently,syntax-aware neural semantic role labeling(SRL)has received much attention.However,most of previous syntax-aware SRL works exploit homogeneous syntactic knowledge from a single syntactic treebank.Considering several high-quality publicly available Chinese syntactic treebanks,this paper proposes to extend graph convolutional networks(GCNs)for encoding heterogeneous syntactic knowledge in the heterogeneous dependency trees and makes a through comparison on various encoding methods to improve SRL performance.This model achieved 84.16 and 85.30 F1 on CPB 1.0 and CONLL-2009 Chinese data sets,respectively,both outperforming the corresponding homogeneous syntax-aware SRL models and significantly improving the performance of previous methods.

关 键 词：语义角色标注 异构句法树库 图卷积网络 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]