The paper introduces a novel model named TransD for knowledge graph embedding, which aims to improve the representation of entities and relations in knowledge graphs. Unlike previous models such as TransE, TransH, and TransR/CTransR, TransD uses two vectors to represent each named symbol object (entity and relation): one captures the meaning, and the other constructs a dynamic mapping matrix. This approach allows TransD to consider the diversity of both entities and relations, leading to more flexible and accurate representations. TransD also avoids matrix-vector multiplication, reducing computational complexity and making it suitable for large-scale graphs. Experimental results on two tasks—triplet classification and link prediction—show that TransD outperforms state-of-the-art methods, including TransE, TransH, and TransR/CTransR. The paper discusses the model's architecture, training objective, and comparisons with related models, highlighting its advantages in handling complex internal correlations in knowledge graphs.The paper introduces a novel model named TransD for knowledge graph embedding, which aims to improve the representation of entities and relations in knowledge graphs. Unlike previous models such as TransE, TransH, and TransR/CTransR, TransD uses two vectors to represent each named symbol object (entity and relation): one captures the meaning, and the other constructs a dynamic mapping matrix. This approach allows TransD to consider the diversity of both entities and relations, leading to more flexible and accurate representations. TransD also avoids matrix-vector multiplication, reducing computational complexity and making it suitable for large-scale graphs. Experimental results on two tasks—triplet classification and link prediction—show that TransD outperforms state-of-the-art methods, including TransE, TransH, and TransR/CTransR. The paper discusses the model's architecture, training objective, and comparisons with related models, highlighting its advantages in handling complex internal correlations in knowledge graphs.