Simple Augmentations of Logical Rules for Neuro-Symbolic Knowledge Graph Completion

Ananjan Nandi,Navdeep Kaur,Parag Singla,Mausam

DOI: https://doi.org/10.18653/v1/2023.acl-short.23

2024-07-02

Abstract:High-quality and high-coverage rule sets are imperative to the success of Neuro-Symbolic Knowledge Graph Completion (NS-KGC) models, because they form the basis of all symbolic inferences. Recent literature builds neural models for generating rule sets, however, preliminary experiments show that they struggle with maintaining high coverage. In this work, we suggest three simple augmentations to existing rule sets: (1) transforming rules to their abductive forms, (2) generating equivalent rules that use inverse forms of constituent relations and (3) random walks that propose new rules. Finally, we prune potentially low quality rules. Experiments over four datasets and five ruleset-baseline settings suggest that these simple augmentations consistently improve results, and obtain up to 7.1 pt MRR and 8.5 pt Hits@1 gains over using rules without augmentations.

Artificial Intelligence,Computation and Language,Information Retrieval

What problem does this paper attempt to address?

The paper primarily aims to address the issue of insufficient rule set coverage in Neuro-Symbolic Knowledge Graph Completion (NS-KGC) models.

Specifically, high-quality and widely covered rule sets are crucial for the success of NS-KGC models, as these rule sets form the basis of all symbolic reasoning. However, existing neural models often struggle to maintain high coverage when generating rule sets, resulting in many queries being unable to trigger symbolic reasoning, thus forcing the model to rely solely on its embedding component. To tackle this problem, the paper proposes three simple rule enhancement techniques:

1. **Inductive Form Transformation of Rules**: Transform each deductive rule into its inductive form.
2. **Inverse Form Rule Generation for Relations**: Supplement each rule with an equivalent rule using the inverse form of its constituent relations.
3. **Random Walk Rule Generation**: Independently generate new high-quality rules through local random walks, followed by Principal Component Analysis (PCA) filtering.

Additionally, the paper proposes a rule filtering method to remove potentially low-quality rules. Experimental results show that these enhancement techniques can significantly improve the performance of NS-KGC models on four datasets, particularly achieving up to 7.1 and 8.5 points improvement in Mean Reciprocal Rank (MRR) and Hits@1 metrics, respectively, compared to baseline models without enhancement techniques.

In summary, this research aims to improve the overall performance of NS-KGC models by enhancing the quality and coverage of rule sets.