Information Gain Calculation for 3 Classes

Question

I have a problem that requires us to use information gain for feature selection but the target class has 3 classes (positive, negative, and neutral) instead of two. As such, I am a little confused about how the calculation for info would work.

For 2 class problems, I know that info(p,n) = 0 if all the objects are in 1 class and there are none in the other class. Would this be the same for 3-class problems? How about if 2 classes have objects and the third class does not?

For example, if I have an attribute split as follows, would the following info(p,n,neu) calculations be correct?

The formula I am using for info is -p/p+n+neulog2(p/p+n+neu)-n/p+n+neulog2(n/p+n+neu)-neu/p+n+neulog2(neu/p+n+neu)

Dept	Positive	Negative	Neutral	Info(p,n,neu)
X	40	34	0	0
Y	60	70	70	0.3969
Z	34	0	36	0
A	68	83	0	0
B	0	50	0	0