Before going into the details of computing outer probabilities we describe their use in obtaining the expected rule counts needed for the E-step in grammar estimation.
Let denote the expected number of uses of production in the derivation of string x. Alternatively, is the expected number of times that is used for prediction in a complete Earley path generating x. Let be the number of occurrences of predicted states based on production along a path .
The last summation is over all predicted states based on production . The quantity is the sum of the probabilities of all paths passing through . Inner and outer probabilities have been defined such that this quantity is obtained precisely as the product of the corresponding of and . Thus, the expected usage count for a rule can be computed as
The sum can be computed after completing both forward and backward passes (or during the backward pass itself) by scanning the chart for predicted states.