电力客服知识图谱的改进研究与模型构建

随着知识图谱在电力服务领域的应用愈加广泛和必要,越来越多的研究者对其进行了深入的研究。其中,基于深度学习的TransE的方法逐渐受到大家的青睐,所以针对其的改进方法也越来越多,例如TransH、TransR/CtransR和TransD等,这些方法均针对"TransE无法解决一对多"等问题,提出了"将TransE模型投影到其他空间"的各种方法。借鉴这些方法的优势,提出了一种基于模糊理论和现有基于深度学习的翻译模型相结合的方法--TransF。在TransF中,为每个元素分别构建两个模糊向量,分别用于构建实体和构建模糊映射,将头尾实体分别映射到与关系构成的模糊空间中。实验结果表明,该方法在训练数据集不大的情况下具有明显优势,并且更符合人类逻辑和实用性的需要。

Research on the Improvement of Power Customer Service Knowledge Graph and Model Construction

As the application of knowledge graph in the field of electric power service becoming more and more extensive and necessary,more and more researchers have conducted in-depth research on it.Among them,the deep learning based on the TransE method is gradually favored by everyone,so there are more and more improved methods for it,such as TransH,TransR/CtransR,and TransD.These methods are all directed to the problem that"TransE cannot solve one to many",Proposed various methods of"projecting TransE model to other spaces".Drawing on the advantages of these methods,a method based on fuzzy theory and the existing deep learning-based translation model is proposed-TransF.In TransF,two fuzzy vectors are constructed for each element,which are used to construct entities and fuzzy maps,respectively.The head and tail entities are mapped to the fuzzy space formed by the relationship.Experimental results show that this method has obvious advantages when the training data set is not large,and it is more in line with the needs of human logic and practicality.