Add-One Smoothing

Add-one smoothing adds 1 to every vocabulary item before computing probabilities.

Suppose the vocabulary is:

tea, code, music

and after I like the counts are:

tea: 2
code: 1
music: 0

After add-one smoothing:

tea: 3
code: 2
music: 1

The total is now 6, so:

P(music | I like) = 1 / 6

Add-one smoothing is easy to understand. It is often too blunt for serious systems, but it teaches the principle clearly.

Exercise

After add-one smoothing in the example above, what is the smoothed count for music?