The Hunt and Black Algorithm

For easy comparison with second-generation techniques, we will assume that we also use diphones in unit selection, bnt, as we will see in Section 16.2.1, a wide variety of other types is possible. As before, the specification is a list of diphone items S = (ii, 52. M Sn), each described by a featme structure. The database is a set of diphone 

In the Hunt and Black framework, unit selection is defined as a search through every possible sequence of units to find the best possible sequence of units. There are several options as to how we define best , but in the original Himt and Black formulation this is defined as the lowest cost, as calculated from two local components. First we have the target cost, T(ui, St), which is a cost or distance between the specification s, and a imit in the database Ui. This cost is calculated from specified values in the featme structure of each unit. Second, we have the join cost, J(ui, M/+i), whichisameasmeofhowwell two units join (low values mean good joins). This is calculated for a pair of imits in the database, again from specified values in the units feature structures. The total combined cost for a sentence is given by 

Section 16.6 will explain how to perform this search, but for now we will simply note that it can be performed as a Viterbi-style search (introduced in Section 15.1.7). 

---

Figure 16.1 A diagram of the Hunt and Black algorithm, showing one particular sequence of units and how the target cost measures a distance between a unit and the specification, and how the join cost measures a distance between the two adjacent units.

Various ways of setting the weights automatically have been proposed. Since the Hunt and Black algorithm is not probabihstic, we cannot directly use the most-standard probabilistic training technique of maximum likelihood. We can, however, investigate similar types of techniques that try to set the parameters so as to generate data that are closest to a set of training data. 

Recall that in the Hunt and Black algorithm (Equations (16.2) and (16.3)) the total cost of one sequence of units from the database U = u, U2. v) with the specification... 

---