TF-IDF (term frequency–inverse document frequency) is a weighting scheme that assigns to each (term, document) pair the product of two factors. The term-frequency factor $\mathrm{tf}(t,d)$ counts (or sublinearly transforms) how often term $t$ occurs in document $d$. The inverse-document-frequency factor $\mathrm{idf}(t)=\log\dfrac{N}{\mathrm{df}(t)}$ down-weights terms that appear in many of the $N$ documents in the corpus, since such terms discriminate poorly. The resulting weight $w_{t,d}=\mathrm{tf}(t,d)\cdot\mathrm{idf}(t)$ is high for terms that are frequent in a document but rare across the corpus.