The present paper is concerned with the automation of the process of conducting literature reviews. Manual reviews are getting more and more difficult as the number of publications increases steeply. Against this backdrop, we investigate the application of deep learning techniques for the automation of the time consuming step of comparing and categorizing large sets of scientific literature. In contrast to prior research, we leverage the potential of the word2vec algorithm that provides a more semantic focus of analysis than common text mining approaches. We evaluate our artifact considering an exemplary document collection comprising 906 articles on Radio Frequency Identification. Our results indicate that our word2vec-based system provides better results than a system based on traditional text mining approaches.