Neural, symbolic and neural-symbolic reasoning on knowledge graphs

Knowledge graph reasoning is the fundamental component to support machine learning applications such as information extraction, retrieval, and recommendation. Since knowledge graphs can be viewed discrete symbolic representations of knowledge, on naturally leverage techniques. However, intolerant ambiguous noisy data. On contrary, recent advances deep have promoted neural graphs, which robust data, but lacks interpretability compared reasoning. Considering advantages disadvantages both methodologies, efforts been made combining two methods. In this survey, we take a thorough look at development symbolic, hybrid graphs. We survey specific tasks — completion question answering explain them in unified framework. also briefly discuss future directions for