This paper presents a novel method for learning structured embeddings of Knowledge Bases (KBs), which are rich sources of structured knowledge. The authors propose a neural network architecture that embeds KB elements into a continuous vector space, preserving the original structure and enhancing the data. This approach allows for easy integration of KB data into machine learning methods for prediction and information retrieval. The method is demonstrated on WordNet and Freebase, and its adaptability to knowledge extraction from raw text is also discussed. The paper highlights the flexibility, compactness, and generalization capabilities of the learned embeddings, making them suitable for various AI applications such as natural language processing, computer vision, and collaborative filtering. The authors also introduce Kernel Density Estimation to estimate the probability density of relations within the embedding space, enhancing the system's ability to infer new relations.This paper presents a novel method for learning structured embeddings of Knowledge Bases (KBs), which are rich sources of structured knowledge. The authors propose a neural network architecture that embeds KB elements into a continuous vector space, preserving the original structure and enhancing the data. This approach allows for easy integration of KB data into machine learning methods for prediction and information retrieval. The method is demonstrated on WordNet and Freebase, and its adaptability to knowledge extraction from raw text is also discussed. The paper highlights the flexibility, compactness, and generalization capabilities of the learned embeddings, making them suitable for various AI applications such as natural language processing, computer vision, and collaborative filtering. The authors also introduce Kernel Density Estimation to estimate the probability density of relations within the embedding space, enhancing the system's ability to infer new relations.