A Knowledge Graph Representation Learning Algorithm Based on Symbolic Semantic Mapping

In order to address the problems of current popular graph representation learning models, the author proposes a knowledge graph representation learning algorithm based on symbol semantic mapping. This algorithm utilizes existing entity relationship data in the knowledge graph to semantically encode symbol combinations through a recurrent neural network and map them to the target symbol. In addition, the model also solves the problem of relationship asymmetry by introducing an inverse relationship mirror for each relationship type, enabling the model to adapt to different types of networks and perform relationship inference tasks. This model is suitable for processing the representation learning tasks of large-scale knowledge maps. The experimental results show that this model outperforms the relevant research work in the knowledge map expansion task and the graph based multi label classification task on the public dataset. This means that this model has potential importance in the application of knowledge graphs and provides an effective tool for the construction and application of knowledge graphs.