Knowledge Representation Learning Based On Multi-source Information Combination

Knowledge Graphs(KGs)construct the existing knowledge into large-scale networks,which are the crystallization of human past experience and intelligence,and now they are playing an increasingly important role in various tasks of artificial intelligence.The gole of Knowledge Representation Learning is to represent the entities and relations in KGs as vectors in low-dimensional space,making the KGs more convenient to be applied to related tasks.In KGs,there are rich contents hidden in the text description information of entity,the hierarchical type information of entity and the topological structure information of graph,and they are effective supplements to the triple information.The combination of these multi-source information is helpful to get better performance in various KGs tasks.In order to make full use of these information,the convolutional neural networks is firstly used to encode entity description.Then two projection matrixs are constructed according to hierarchical type information to projects entity vectors and entity description vectors into specific relation space to constrain their semantic information.After that,the graph attention mechanism is introduced to fuse the topological structure information of graph and calculate the influence of different adjacency points on entities.Meanwhile,the multi-hop relationship information between entities is calculated to further solve the problem of data sparsity.Additionally,in the process of training,the generation adversarial networks is applied to generate high-quality negative sample triples to give full play to the ability of the model.Finally,a decoder is employed to capture the global information between different dimensions.The experiment results of link prediction and triple classification show that the combined model can make good use of multi-source heterogeneous information beyond triples,so it obtains better results than other baseline models.