Smoothing
From Cohen Courses
Jump to navigationJump to searchFrom Wikipedia:
In statistical language modeling, in a bag of words model for example, the data consists of the number of occurrences of each word in a document. Smoothing allows the assignment of non-zero probabilities to words which do not occur in the sample. From a Bayesian point of view, this corresponds to the expected value of the posterior distribution of words, using a Dirichlet distribution with parameter α as a prior.