Given a term-document matrix \(X\), let row \(i\) represent document \(i\); column \(j\) represent term \(j\).

\[ \text{occurence} = \log (1 + X_{ij}) \]

Taking the SVD of \(X\), each component are:

\(u_i\): Similar or related terms (i.e., genre) \(v_i\): Frequent terms