Amir Hosein Baqinejad
Interested Fields
Deep Learning, SE
Title of Thesis
Temporal Knowledge Graph Embedding with Concentrating on Context Information
Description of Thesis
Knowledge graph embedding has been gaining more attention, especially in knowledge graph completion tasks. With this method, triplets of knowledge graph that contain head entity, relation, and tail entity are embedded in a continuous vector space and we can infer new links between entities. Temporal knowledge graphs, in addition to the head entity, relation, and tail entity, in each triplet, have another element that shows the time that the triplet has occurred or has been valid. With this new element, we can increase the accuracy of knowledge graph embedding and knowledge graph completion. We know that an entity can have some roles, as an example a president can have a political role, and also at the same time, he/she can be a member of a family. Also, a relation can have different meanings according to its head and tail entities. By paying attention to the paths in the knowledge graph, we can know about the role of an entity and the meaning of a relation. This method is called paying attention to the context. This method also can help us increase the accuracy of knowledge graph embedding and knowledge graph completion. We want to combine the two aforementioned ideas to use the advantages of both of them.
Contact Information
Skype: ambaqinejad
ambaqinejad@gmail.com