基于事件认知的知识图谱动态推理机制研究

知识图谱具有数据规模大、语义丰富、结构友好和流式、动态的特点，通过人机友好的方式让知识被机器理解和运用，并将其用于推理以实现智能认知。随着研究的不断深入，推理过程中出现的图谱结构多变、异构关系复杂、逻辑关系多样等困难，对知识图谱的动态推理提出了新的挑战。基于此，本项目将事件的动态演变作为知识推理的基本线索，从知识推理模式改变、时效增强、语义转换等视角出发，研究事件驱动的知识图谱重构与动态融合、面向时序关联的知识图谱表示学习、基于拓扑结构动态感知的知识推理以及基于事件逻辑联合感知的知识推理等关键机制，以提升知识图谱推理的实时性、置信度和可解释性，并为个性化推荐、智能问答、语义搜索等提供技术支撑。本项选题具有先进性，研究具有重要的理论意义与实用价值，研究成果可拓宽知识图谱知识推理的研究领域，为知识图谱深度分析与应用推广奠定基础。

Knowledge Graph (KG), which is not only designed for making machine understand and use knowledge in a human-computer friendly way but also aiming at using it for deep reasoning and achieving intelligent cognition, has the features of large-scaled, semantic-riched, well-structured, streamlined and dynamic. As further research continues, several difficulties such as intricate graph structure, complex heterogeneous relevance and diverse logical relations have appeared in reasoning process, which brings new challenges to dynamic reasoning in KG. Based on these, from the perspectives of changes in knowledge reasoning models, enhanced timeliness and semantic transformation, this project, leveraging the dynamic evolution of events as a basic clue for knowledge reasoning, investigate key mechanisms such as event-driven KG reconstruction and dynamic fusion, temporal-oriented KG representation learning, knowledge reasoning based on dynamic perception of topology and joint perception of event logic, to boost the real-time ability, confidence and explainablity and provide technical supports for personalized recommender systems (RS), intelligent question answering (QA) and semantic-enhanced search engine (SE). The selection of this project is extremely advanced and research based on this has essential theoretical significance and practical value. The research results can broaden the field of knowledge reasoning in KG and firmly lay a foundation for deep analysis and applications of KG.