Uncertain Knowledge Graph Completion with Rule Mining

Yilin Chen, Tianxing Wu, Yunchang Liu, Yuxiang Wang, Guilin Qi

Abstract: To model the uncertainty within knowledge graphs (KGs), existing studies define uncertain knowledge graphs (UKGs), which assign a confidence score to each triple to measure its likelihood of being true and make more precisely downstream tasks such as reasoning and decision making possible. Since KGs usually suffer from the problem of incompleteness, methods of rule mining and reasoning for knowledge graph completion are extensively studied due to their excellent interpretability. However, previous methods are all conducted under deterministic scenarios, neglecting the uncertainty of knowledge, making them unable to be directly applied to UKGs. In this paper, we propose a new framework on uncertain knowledge graph completion with rule mining. The framework is composed of a rule mining model and a confidence prediction model. The rule mining model applies an encoder-decoder network transformer to take rule mining as a sequence-to-sequence task to generate rules. It models the uncertainty in UKGs and infer new triples by differentiable reasoning based on TensorLog with mined rules. The confidence prediction model uses a pre-trained language model to predict the triple confidence given the rules mined. Experiments show that our models significantly outperform various baselines in different evaluation metrics on link prediction and confidence prediction, respectively.