Semantic Evidence Research Articles

How Does Knowledge Graph Embedding Extrapolate to Unseen Data: A Semantic Evidence View

Knowledge Graph Embedding (KGE) aims to learn representations for entities and relations. Most KGE models have gained great success, especially on extrapolation scenarios. Specifically, given an unseen triple (h, r, t), a trained model can still correctly predict t from (h, r, ?), or h from (?, r, t), such extrapolation ability is impressive. However, most existing KGE works focus on the design of delicate triple modeling function, which mainly tells us how to measure the plausibility of observed triples, but offers limited explanation of why the methods can extrapolate to unseen data, and what are the important factors to help KGE extrapolate. Therefore in this work, we attempt to study the KGE extrapolation of two problems: 1. How does KGE extrapolate to unseen data? 2. How to design the KGE model with better extrapolation ability? For the problem 1, we first discuss the impact factors for extrapolation and from relation, entity and triple level respectively, propose three Semantic Evidences (SEs), which can be observed from train set and provide important semantic information for extrapolation. Then we verify the effectiveness of SEs through extensive experiments on several typical KGE methods. For the problem 2, to make better use of the three levels of SE, we propose a novel GNN-based KGE model, called Semantic Evidence aware Graph Neural Network (SE-GNN). In SE-GNN, each level of SE is modeled explicitly by the corresponding neighbor pattern, and merged sufficiently by the multi-layer aggregation, which contributes to obtaining more extrapolative knowledge representation. Finally, through extensive experiments on FB15k-237 and WN18RR datasets, we show that SE-GNN achieves state-of-the-art performance on Knowledge Graph Completion task and performs a better extrapolation ability. Our code is available at https://github.com/renli1024/SE-GNN.

SEE: structured representation of scientific evidence in the biomedical domain using Semantic Web techniques.

BackgroundAccounts of evidence are vital to evaluate and reproduce scientific findings and integrate data on an informed basis. Currently, such accounts are often inadequate, unstandardized and inaccessible for computational knowledge engineering even though computational technologies, among them those of the semantic web, are ever more employed to represent, disseminate and integrate biomedical data and knowledge.ResultsWe present SEE (Semantic EvidencE), an RDF/OWL based approach for detailed representation of evidence in terms of the argumentative structure of the supporting background for claims even in complex settings. We derive design principles and identify minimal components for the representation of evidence. We specify the Reasoning and Discourse Ontology (RDO), an OWL representation of the model of scientific claims, their subjects, their provenance and their argumentative relations underlying the SEE approach. We demonstrate the application of SEE and illustrate its design patterns in a case study by providing an expressive account of the evidence for certain claims regarding the isolation of the enzyme glutamine synthetase.ConclusionsSEE is suited to provide coherent and computationally accessible representations of evidence-related information such as the materials, methods, assumptions, reasoning and information sources used to establish a scientific finding by adopting a consistently claim-based perspective on scientific results and their evidence. SEE allows for extensible evidence representations, in which the level of detail can be adjusted and which can be extended as needed. It supports representation of arbitrary many consecutive layers of interpretation and attribution and different evaluations of the same data. SEE and its underlying model could be a valuable component in a variety of use cases that require careful representation or examination of evidence for data presented on the semantic web or in other formats.