Recently, Linked Open Data (LOD) has experienced an exponential growth via publishing huge volume of datasets on the Web. This vast amount of information needs to be searched, queried, and interlinked easier than before. It is recommended that potential data publishers provide recapitulative information about their datasets published on the Web. This information, which functions as metadata, will facilitate those datasets to be discovered easily. As it is not always the case, we are faced with a large number of datasets without a proper profile, leading to a high demand for different data profiling techniques. In this paper, we focus on RDF dataset profiling utilizing unsupervised machine learning techniques, namely knowledge based topic modeling. We also investigate the use of Wikipedia categories to represent the topics identified in an RDF dataset. In the proposed model, we extract a number of representative topics for an RDF dataset and annotate them with Wikipedia categories. The union of the assigned categories serves as a profile of the dataset, in a sense that it provides an overall characterization of the content of the dataset.