大模型知识引导的复合多注意力文档级关系抽取方法
|
|
竹志超,李建强,齐宏智,赵青,高齐,李思颖,蔡嘉怡,沈金炎
|
Large model knowledge-guided composite multi-attention method for document-level relation extraction
|
|
Zhichao ZHU,Jianqiang LI,Hongzhi QI,Qing ZHAO,Qi GAO,Siying LI,Jiayi CAI,Jinyan SHEN
|
|
| 表 4 LKCM与基线模型的统计显著性分析结果 |
| Tab.4 Statistical significance analysis results between LKCM and baseline models |
|
| 模型 | p-value | | P | R | F1 | | LKCM/RoBERTa-base | 3.88×10−4 | 2.86×10−4 | 3.21×10−4 | | LKCM/SSAN | 1.70×10−3 | 3.22×10−3 | 1.90×10−3 | | LKCM/GCGCN | 3.09×10−4 | 2.51×10−2 | 2.88×10−3 | | LKCM/GLRE | 1.63×10−2 | 4.29×10−4 | 3.74×10−3 | | LKCM/GRACR | 3.14×10−3 | 3.82×10−3 | 3.51×10−3 | | LKCM/DISCO | 7.25×10−5 | 6.68×10−3 | 6.00×10−3 | | LKCM/K-BiOnt | 3.46×10−3 | 1.24×10−2 | 1.65×10−2 | | LKCM/KIRE | 6.48×10−4 | 3.50×10−3 | 7.90×10−4 | | LKCM/KRC | 2.19×10−3 | 6.23×10−3 | 3.06×10−4 | | LKCM/GECANet | 1.58×10−4 | 1.32×10−2 | 3.01×10−3 |
|
|
|