Example: nlp_speech method.

A more interesting example is provided by the following method which not only takes into account acoustic scores, but also the number of skip edges and category edges. Weights are expressed as $\langle c_1, c_2, c_3\rangle$, where c₁ is the number of skips, c₂ is the number of categories, and c₃ is the acoustic score.

We define $\mbox{\it ini\/} = \langle 0, 0, 0\rangle$ and $\mbox{\it uw\/}$ is defined as follows.

$$\mbox{\it uw\/}(\langle c_1, c_2, c_3\rangle,(v_i,v_j,w,a,q)... ...c_3+a\rangle & \mbox{for stopping edges}\\ \end{array}\right.$$ (32)

Finally, we define the ordering on such tuples:

$$\langle c_1, c_2, c_3\rangle \prec \langle c'_1, c'_2, c'_3\... ...nd } c_2 = c'_2 \mbox{ and } c_3 < c'_3\\ \end{array}\right.$$ (33)