Knowledge graph completion method based on quantum embedding and quaternion interaction enhancement

Knowledge graphs (KG) are used for many downstream tasks in artificial intelligence (AI). However, owing to accuracy issues associated with information extraction, KGs are often incomplete. This has led to the emergence of knowledge graph completion (KGC) tasks. Their purpose is to learn known facts to infer the missing entities in triples. Traditional embedding-based methods usually only focus on the information of individual triples and do not use the deep logical relationships of the KG. In this study, we propose a new KGC method referred to as QIQE-KGC. It uses quantum embedding and quaternion space interaction to capture the external logical relationship between triples in a KG and enhance the connection between entities and relations within a single triple to model and represent the KG. The proposed QIQE-KGC model can capture richer logical information and has more powerful and complex relationship modeling capabilities. Extensive experimental results using QIQE-KGC on 11 datasets demonstrate that the model achieves outstanding performance. Compared to the baseline models, QIQE-KGC produced the best results on most datasets.