Time-Aware Graph Embedding: A Temporal Smoothness and Task-Oriented Approach

Knowledge graph embedding, which aims at learning the low-dimensional representations of entities and relationships, has attracted considerable research efforts recently. However, most knowledge embedding methods focus on structural relationships in fixed triples while ignoring temporal information. Currently, existing time-aware only factual plausibility, smoothness, models interactions between a fact its contexts, thus can capture fine-granularity relationships. This leads to limited performance related applications. To solve this problem, article presents Robustly Time-aware Graph Embedding (RTGE) method by incorporating smoothness. Two major innovations our are presented here. At first, RTGE integrates measure smoothness process embedding. Via proposed additional smoothing factor, preserve both information evolutionary patterns given graph. Secondly, provides general task-oriented negative sampling strategy associated with temporally aware information, further improves adaptive ability algorithm plays an essential role obtaining superior various tasks. Extensive experiments conducted multiple benchmark tasks show that increase entity/relationship/temporal scoping prediction