The Web of Data introduces issues in storing, representing, managing, and querying the massive amount of data with additional information i.e., meta-knowledge. These issues exist because when introducing meta-knowledge the graph size, the number of statements, and the query response time increase, and some redundancies are created. We propose a new approach called Labeled k-partite Graph data model (LKG) to overcome some of these issues. Our approach uses MSPARQL as a query language, an extension of SPARQL. Experiments were conducted using the SPARQL Performance Benchmark (SP 2 Bench) dataset (without meta-knowledge) and, the Bio-medical Knowledge Repository (BKR) dataset and Gov-track dataset (with meta-knowledge). For the former dataset, the LKG approach is compared with the RDF directed labeled graph, and for the latter, LKG is compared with the state-of-the-art approaches Singleton Property, RDF Reification, Named Graph and PaCE. The experiments analyzed the meta-knowledge representation in terms of number of edges, number of statements, storage space and redundancy creation. The results show that the LKG model performs better in terms of query response time, query length, and in terms of meta-knowledge representation, which gives an advantage over storage management, graph data management and gives also a faster information retrieval.