关键词: 知识图谱, 多关系推理, 近似推理, 不确定性, 卷积神经网络

Abstract: In response to the issue that approximate reasoning on multiple logical relations could not be performed by the uncertain embedding model of a large-scale Knowledge Graph(KG), UDConEx(Uncertainty DistMult and Complex Convolution Embedding) was proposed, which was a multi-relation approximate reasoning model based on uncertain knowledge graph embedding. Firstly, the characteristics of the DistMult(Distance Multiplicative) model and the ComplEx(Complex Embedding) model were combined within the UDConEx model, bestowing upon it the capability to infer symmetric and asymmetric relationships. Subsequently, Convolutional Neural Network(CNN) was employed by the model to capture the interactive information in the uncertain KG, thereby enabling it to possess the ability to reason about inverse and transitive relationships. Lastly, the neural network was employed to carry out confidence learning of uncertain KG information, enabling the UDConEx model to perform approximate reasoning within the uncertain KG embedding space. The experimental results on three standard data sets of CN15k, NL27k, and PPI5k show that compared with the newly proposed MUKGE model, the Mean Absolute Error(MAE) of confidence prediction is reduced by 1.25, 1.46, and 1.62 respectively; in the task of relation fact ranking, the linear-based Normalized Discounted Cumulative Gain(NDCG) improved by 5.8% and 2.6% in the CN15k and NL27k datasets, respectively; in the multi-relation approximate reasoning task, it is verified that the model has the approximate reasoning ability of multiple logical relationships. The inability of traditional embedding models to predict confidence is compensated for by the UDConEx model, which achieves approximate reasoning for multiple logical relationships and offers enhanced accuracy and interpretability in uncertainty KG reasoning capabilities.

Key words: Knowledge Graph(KG), multi-relation reasoning, approximate reasoning, uncertainty, Convolutional Neural Network(CNN)