Research On Knowledge Graph Embedding Methods In Dynamic Environment

Knowledge graph(KG)embedding encodes the entities and relations from a KG into lowdimensional vector spaces to support various applications such as KG completion,question answering,and recommender systems.In real world,knowledge graphs(KGs)are dynamic and evolve over time with addition or deletion of triples.However,most existing models focus on embedding static KGs while neglecting dynamics.To adapt to the changes in a KG,these models need to be re-trained on the whole KG with a high time cost,while not able to update KG embeddings in a more effective way.In this thesis,to tackle the problem mentioned above,we propose a new context-aware Dynamic Knowledge Graph Embedding(DKGE)method which supports efficient embedding learning.DKGE introduces two different representations(i.e.,knowledge embedding and contextual element embedding)for each entity and each relation,in the joint modeling of entities and relations as well as their contexts,by employing two attentive graph convolutional networks,a gate strategy,and translation operations.This not only helps improve the performance of embedding learning,but also effectively helps limit the impacts of a KG update in certain regions,not in the entire graph,so that DKGE can rapidly acquire the updated KG embedding by a proposed online learning algorithm.The main contributions of this thesis are as follows:1)Proposing a context-based knowledge graph embedding model,which gets high-quality embeddings of entities and relations by encoding the contexts of entities and relations in into embeddings and jointly training them with entities' and relations' embeddings.2)Proposing an online learning algorithm of KG embedding,which is designed based on the abovementioned context-based KG embedding model and able to get new high-quality embeddings efficiently after a KG update.3)Performing experiments on real-world datasets and evaluating the proposed method using widely-used criteria,the results show the abovementioned context-based KG embedding model performs better than other state-of-the-art models and the online learning algorithm achieves better performance in both effectiveness and efficiency(especially in efficiency).Researching on KG embedding methods in dynamic environment can help apply KG embedding techniques more easily and efficiently in dynamic environment and expand lots of applications based on KG and KG embedding to dynamic scenarios.This is significant not only for the development of KGs but also for the development of various applications based on KG and KG embedding.