ABSTRACT
In the earlier sections of this chapter, the feature overlapping rules were constructed based only on the information about the phoneme or its constituents' identity in the utterance. It is well established [Brow 89, Church 87, Cole 98, Golds 90] that a wealth of linguistic factors beyond the level of phoneme, in particular prosodic information (syllable, morpheme, stress, utterance boundaries, etc.), directly control the low-level feature overlapping. Thus, it is desirable to use such high-level linguistic information to control and to constrain feature overlapping effectively. As an example, in pronouncing the word display, the generally unaspirated [p] is constrained by the condition that an [s] precedes it in the same syllable onset. On the other hand, in pronouncing the word displace, dis is a morphological unit of one syllable and the [p] in the initial position of the next syllable subsequently tends to be aspirated.
