[论文解读] Learning First-Order Rules with Relational Path Contrast for Inductive Relation Reasoning

Yudai Pan, Jun Liu+4 more|arXiv (Cornell University)|Oct 17, 2021

Advanced Graph Neural Networks参考文献 29被引用 7

一句话总结

该论文提出RPC-IR，一种基于图卷积网络的新型归纳关系推理方法，通过关系路径对比学习学习可解释的一阶规则。通过提取关系路径并应用对比策略增强自监督学习，RPC-IR在三个归纳数据集上实现了最先进性能，同时显式表示逻辑规则以实现可解释性。

ABSTRACT

Relation reasoning in knowledge graphs (KGs) aims at predicting missing relations in incomplete triples, whereas the dominant paradigm is learning the embeddings of relations and entities, which is limited to a transductive setting and has restriction on processing unseen entities in an inductive situation. Previous inductive methods are scalable and consume less resource. They utilize the structure of entities and triples in subgraphs to own inductive ability. However, in order to obtain better reasoning results, the model should acquire entity-independent relational semantics in latent rules and solve the deficient supervision caused by scarcity of rules in subgraphs. To address these issues, we propose a novel graph convolutional network (GCN)-based approach for interpretable inductive reasoning with relational path contrast, named RPC-IR. RPC-IR firstly extracts relational paths between two entities and learns representations of them, and then innovatively introduces a contrastive strategy by constructing positive and negative relational paths. A joint training strategy considering both supervised and contrastive information is also proposed. Comprehensive experiments on three inductive datasets show that RPC-IR achieves outstanding performance comparing with the latest inductive reasoning methods and could explicitly represent logical rules for interpretability.

研究动机与目标

• 为解决归纳知识图谱推理中实体无关的关系语义挑战。
• 克服因潜在规则稀缺导致子图中监督不足的问题。
• 实现具有可解释性的、逻辑性的一阶规则的归纳推理。
• 通过学习与特定实体无关的关系序列语义，提升对未见实体的泛化能力。
• 通过对比关系路径训练，整合监督学习与自监督学习。

提出的方法

• RPC-IR在每个子图中提取预定义长度内的实体间关系路径。
• 利用图卷积网络（GCN）表示这些路径，以学习与实体无关的关系语义。
• 通过构建正样本与负样本关系路径对，引入对比学习策略，以增强自监督信号。
• 模型联合优化来自标注三元组的监督损失与来自正/负路径对的对比损失。
• 从高置信度关系路径中提取一阶规则，置信度分数反映推理的可靠性。
• 通过从子图结构中泛化，无需对未见实体重新训练，支持归纳推理。

实验结果

研究问题

• RQ1 关系路径对比学习能否提升知识图谱中的归纳关系推理？
• RQ2 RPC-IR在多大程度上能从子图路径中学习到与实体无关的关系语义？
• RQ3 联合监督与对比训练在归纳基准上的性能提升效果如何？
• RQ4 该模型能否生成可解释的一阶规则以解释推理预测？
• RQ5 超参数λ₁和λ₂在不同数据集上的表现受何影响？

主要发现

• RPC-IR在三个完全归纳数据集（WN18RR、FB15K-237和NELL-995）上达到最先进性能。
• 在FB15K-237_v1上，RPC-IR在λ₁=0.8且λ₂=0.8时达到AUC-PR为0.82，优于先前的归纳方法。
• 该模型学习到显式的一阶规则，如'partOf(X,Z) ∧ locatedIn(Z,Y) → liveIn(X,Y)'，在FB15K-237_v1上置信度高达1.00。
• 案例研究显示，置信度<0.01的规则被过滤，表明规则提取具有可靠性。
• 超参数分析表明，λ₁与λ₂的最优值因数据集而异，FB15K-237上平衡监督效果最佳。
• 对比策略显著提升性能，有效缓解了子图中监督稀疏的问题。