基于异质图卷积神经网络的论点对抽取模型

刘议丹,朱小飞,尹雅博

Heterogeneous graph convolutional neural network for argument pair extraction

Yidan LIU,Xiaofei ZHU,Yabo YIN

表 2 HGCN-APE在RR-passage和RR-submission-v2数据集上的性能对比

Tab.2 Comparison of performance of HGCN-APE on RR-passage and RR-submission-v2 dataset

数据集 模型 论点预测 句子对预测 论点对预测 P/% R/% F1/% P/% R/% F1/% P/% R/% F1/% RR-submission-v2 MT-H-LSTM-CRF 70.74 69.46 70.09 52.05 46.74 49.25 27.24 26.00 26.61 MLMC 69.53 73.27 71.35 60.01 46.82 52.60 37.15 29.38 32.81 MGF 70.40 71.87 71.13 — — — 34.23 34.57 34.40 MRC-APE-Bert 73.36 68.35 70.77 — — — 42.26 34.06 37.72 HGCN-APE 71.86 71.80 71.83 66.12 59.40 62.58 42.70 36.05 39.09 RR-passage MT-H-LSTM-CRF 71.85 71.01 71.43 54.28 43.24 48.13 30.08 29.55 29.81 MLMC 66.79 72.17 69.37 62.49 42.33 50.53 40.27 29.53 34.07 MGF 73.62 70.88 72.22 — — — 38.03 35.68 36.82 MRC-APE-Bert 66.81 69.84 68.29 — — — 34.70 35.53 35.51 HGCN-APE 72.50 71.61 72.05 67.68 59.25 63.18 45.76 38.32 41.71