The add-k smoothed bigram probability is given by
$$P^*_{Add-k}(w_n|w_{n-1}) = \frac{C(w_{n-1}w_n) + k}{C(w_{n-1}) + kV}$$

University of Michigan - Ann Arbor

Add-K Smoothing adds a fractional count k to each count instead of adding one, it moves less probability mass from the seen to the unseen events than add-one smoothing. Add-k smoothing requires a method for choosing k; this can be done, for example, by optimizing on a development set.

Add-K Smoothing

An on-going but a helpful book resource about NLP
https://web.stanford.edu/~jurafsky/slp3/

Learn Before

Related