A Comparison Study of Direct Inference and Knowledge Compensating Generalized Inference as Multidisciplinary for Medical Knowledge Graph

Knowledge graph has drawn increasingly attention in medical artificial intelligence in recent years. In the task of medical knowledge inference, most existing approaches are focusing on single disciplinary, leaving the idea of multidisciplinary behind. In consideration of the distinct advantages of multidisciplinary in clinical decision, simulating multidisciplinary by medical knowledge graph alignment can also prompt the medical knowledge inference. To introduce the idea of multidisciplinary into medical knowledge inference, we design a preliminary pipeline called knowledge compensating generalized inference, which consists of knowledge alignment and embedding based link prediction. In the comparison study, the knowledge compensating generalized inference outperforms the direct inference of single disciplinary, showing a promising potential in medical artificial intelligence applications.