Clustering and latent semantic indexing (LSI) are the most common data analysis in text mining. Yet, usually these tasks are discussed separately even though both involve computing the same factors. In this book, we will treat these two seemingly different concepts as two aspects of the same mathematical formula. The standard methods in clustering and LSI produce mixed signed factors which are unintuitive since most real datasets are nonnegative. Hence, it is natural to consider the using of nonnegative matrix factorizations which can offer more interpretable results. The discussions in this book are both theoretical and practical since we give mathematical proofs for some important results and accompany our algorithms with working codes in Matlab/Octave scripts. Thus, both scholarly and practical readers can benefit from this book.