在大数据发展所造就的数字经济时代，商务大数据融合与分析的关键技术与方法成为最显著也亟待突破的难题。本项目以阿里藏经阁（知识引擎）研究计划为依托背景，以阿里电商平台积累的多源异构商务大数据为基础，重点围绕商务大数据所面临的知识多样性和逻辑复杂性、全景式知识图谱融合、高动态高时效的场景化决策管理等问题，开展面向商务大数据的知识图谱引擎构建方法和关键技术研究。具体以面向符号与向量相结合的知识图谱表示学习框架为技术基础，开展基于知识表示学习的大规模知识图谱融合、场景化知识图谱补全与推理分析等方面的关键技术研究，涉及的关键科学问题包括：复杂知识结构的表示学习、全景式深度语义融合和高动态高时效的可微分知识推理。并结合多个真实的商务场景加以应用（如新零售、跨境电商等），为多边市场主体的决策与管理进行知识赋能，包括精准捕捉并智能识别消费趋势、基于多图谱融合的货品重组、智能导购链路和智能管控等。

In the big data era, business big data fusion and analysis become the most obvious and urgent problems. Based on the Alibaba's Sutra Pavilion Research project and the multi-source heterogeneous business big data accumulated by Alibaba's e-commerce platform, this program takes the key technologies for knowledge graph engine construction for business big data as the overall research goal, particularly focusing on solving the problem of panoramic knowledge graph fusion and semantic reasoning analysis. This program proposes to study logic-assissted knowledge graph embedding framework which tries to integrate the advantages of both symbolic and vector-based knowledge representation. Based upon the novel knowledge embedding framework, this program carries out the key technology research on large-scale business data fusion, highly dynamic knowledge inference and semantic analysis, and combines many real business scenarios (such as new retail, oversea e-commerce, etc.).