Deterministic synthesis methods

Given a representation of intonational form predicted from text, we next attempt to synthesize an FO contour from this. In this section we will examine two rule based deterministic proposals for doing this, the first developed for use the with original Pierrehumbert model and the second, developed more specifically for use with ToBI. 

Intonational event context resultant time position 

Intonational event Context Resultant time position Resultant pitch range position 

---

Finally, the methods of tasks A, C, D are non-deterministic, but finite. This means that choice-points are created there, and that the made selections may be reconsidered later, either because synthesis fails, or because synthesis succeeds and the specifier wants more algorithms. Only Task A could possibly fail, namely if there is no parameter of an inductive type. Do not mix up non-deterministic synthesis and a non-deterministic synthesized algorithm. The latter would feature either mutually nonexclusive cases or predicates that are non-deterministic. The logic algorithm synthesized at Step 2 is deterministic, because the two cases are mutually exclusive by construction, and because the introduced predicates are deterministic. 

Prominence prediction by deterministic means is actually one of the most successful uses of non-statistical methods in speech synthesis. This can be attributed to a number of factors, for example the fact that the rules often don t interact or the fact that many of the rules are base on semantic features (such that even if we did use a data driven technique we would still have to come up with the semantic taxonomy by hand). Sproat notes [410] that statistical approaches have had only limited success as the issue (especially in compound noun phrases) is really one of breadth and not modelling regardless of how the prominence algorithm actually works, what it requires is a broad and exhaustive list of examples of compound nouns. Few complex generalisations are present (what machine learning algorithms are good at) and once presented with an example, the rules are not difficult to write by hand. 

The paper then focusses on 2 recent developments, namely response surface methods (RSM) and component mode synthesis (CMS). The former is a strategy to approximate the real structural behaviour as modelled by repeated finite element calculations by a response surface model that interpolates for structural response between well-chosen settings of uncertain parameters. The latter is an established theory for modelling structural dynamics using superelements. The paper extends the deterministic CMS to models with fuzzy uncertain parameters in the superelements. 

ProbabiUstic methods are based on time series analysis and synthesis. They combine deterministic and statistical analysis, and they synthesize a time (or space) series of stochastic variables and the effects of a limited number of data. It is assumed that the series represents both definable causes and an unknown number of stochastic causes, and that the stochastic causes are reasonably independent. With these methods, jumps, trends and outliers of the data set can be adequately taken into account. It is emphasized that the data used in probabilistic evaluations are based on actual measurements or variables. As with deterministic methods, probabilistic methods should be used in conjunction with engineering judgement when it is feasible, they should be checked by the use in parallel of a simplified deterministic analysis. 

---