Graph structure data is created in many real applications. Graphs represent data by displaying entities (nodes) and their relationships (edges). Several methods are available for effective searching in these networks when graphs are large and heterogeneous and when the patterns are complex and noisy. Graph embedding provides an effective solution to this problem. In graph embedding, the nodes in the graph are evaluated based on their significance. There will be a wide range of graph-related works presented, including classifications in latent space and neural network representations. A variety of methods can be used to integrate node features and label information. In heterogeneous networks, the learned latent space becomes closer to similar nodes. Due to their rich semantic information and heterogeneity, there are challenges associated with analyzing heterogeneous graphs embedded in latent space. Most of the existing methods consider nodes with a single label. Therefore, these methods cannot manage a network of multilabel nodes that describes complex concepts associated with samples. It improves the embedding of heterogeneous networks by combining labels and modeling heterogeneous networks. This is a challenging task. This article presents a node embedding method based on a graph neural network (GCN) to address these issues. By incorporating label information into heterogeneous network embedding, relationships between labels are improved. In comparison to embedded heterogeneous network algorithms, the presented method performed well with real-world data. Graph embedding is discussed in the first section of this paper. The final portion of this paper discusses the application of multi-label embedding to heterogeneous graphs and future research directions.