SSP: Semantic Space Projection for Knowledge Graph Embedding with Text Descriptions

摘要: Knowledge representation is an important, long-history topic in AI, and there have been a large amount of work for knowledge graph embedding which projects symbolic entities relations into low-dimensional, real-valued vector space. However, most methods merely concentrate on data fitting ignore the explicit semantic expression, leading to uninterpretable representations. Thus, traditional limited potentials many applications such as question answering, entity classification. To this end, paper proposes method \textbf{(KSR)}, imposes two-level hierarchical generative process that globally extracts aspects then locally assigns specific category each aspect every triple. Since both categories are semantics-relevant, collection treated Extensive experiments justify our model outperforms other state-of-the-art baselines substantially.