In the heterogeneous network formed by many online systems, such as social search and academic search, name entity disambiguation is a key problem and also a very tricky problem.Name entity disambiguation is designed to resolve the ambiguity caused by the author of the same name.This paper proposes a disambiguation model of named entity based on meta-path heterogeneous network embedding for the disambiguation problem of DBLP database.This model firstly extracts the author information, title and conference name of the article, then generates the publication journal name and the publication title matrix through word2vec model tool for word embedding.Then it is input into the GRU network for training. The GRU network is a variant of LSTM in the neural network of the computer, and then a PHNet is constructed for random walk and the meta path-based walk is used to capture the relationship between different types of nodes.Finally, the weighted heterogeneous network is embedded to realize name disambiguation.The raw data set of the experiment is the open data set of the large-scale online academic search system DBLP, and the experimental results show that the algorithm has good performance in accuracy, recall rate and other indicators.