Prosodic Form

Informally we can describe prosody as the part of human communication which expresses emotion, emphasises words, shows speaker attitude, breaks a sentence into phrases, governs sentence rhythm and controls the intonation, pitch or tune of the utterance. This chapter describes how to predict prosodic form from the text while Chapter 9 goes on to describe how to synthesize the acoustics of prosodic from these form representations. In this chapter we first we introduce the various manifestations of prosody in terms of phrasing, prominence and intonation. Next we go on to describe how prosody is used in communication, and in particular explain why this has a much more direct affect on the final speech patterns than with verbal communication. Finally we describe techniques for predicting what prosody should be generated from a text input. 

To some extent prosody can be seen as a parallel communication mechanism to verbal language. There we had meaning, form and signal, and broadly speaking we can use 

---

Another passage, whose prosodic form is different and therefore is likely to derive from a different text, refers to inner essences circulated and joined by the adept in meditation ... 

Firstly, let us separate the issue of generating a signal which contains a particular prosodic effect (for example surprise, or accentuation of a particular word) from the issue of deciding how to do this automatically from the text. The issue of the realisation of a particular prosodic form is by no means easy, but it is certainly more tractable than the second issue. In essence it is no different fi om other parts of the synthesis problem in that we can collect data with this effect, study how supra-segmental features vary with respect to this effect and so on. 

Prosody prediction Attempting to predict a prosodic form for each utterance from the text. This includes... 

Taking the second issue first we see that in every area of prosody there is considerable disagreement as to how to represent prosodic form. The problem isn t so much that researchers disagree... 

Given these difficulties with labelling schemes and models, it is worth asking whether we need a model of prosodic form at all perhaps it would be possible to simply define a mapping from the text or semantics to the acoustics. This viewpoint certainly has some merit, but we should be wary of dismissing the idea of prosodic form entirely. A brief digression into the theoretical history of prosody may help explain why. 

It is usual to consider three basic components of prosodic form phrasing, prominence and intonation. 

Very little agreement exists on how to represent prosodic form, which makes the design of practical algorithms more difficult than in other areas of TTS. 

Most popular machine learning algorithms have been applied to prosodic prediction. The results are often similar, and difficult to improve upon because of the inherent difficulties in representing prosodic form and under specification in the text. 

This chapter is concerned with the issue of synthesising acoustic representations of prosody. The input to the algorithms described here varies but in general takes the form of the phrasing, stress, prominenee and discourse patterns which we introdueed in Chapter 6. Hence the complete process of synthesis of prosody ean be seen as one where we first extract a prosodic form representation from the text, described in Chapter 6, and then S5mthesize an acoustic representation of this form, deseribed here. 

We now turn our attention to this issue of intonation synthesis itself. In Chapter 6 we described a variety of techniques for generating the prosodic form representations of phrasing and prominence from text. We only made passing reference to the issue of generating intonational form fi-om text as this really required a thorough discussion on the issue of intonational form first. With this job now done, we can describe techniques for both generating intonational form descriptions fi om text and for generating FO controls from intonational form. 

The history of intonation research is full of attempts at finding a system of prosodic form which could act as the equivalent of phonemes in the verbal component of language... 

Some models, such as SFC, do away even with notions of explicit prosodic form and learn the all the processes and representations of intonation from data. 

In the remainder of this chapter we shall concentrate on prosody prediction, the term given to the task of generating the prosodic form from the text. In Chapter 9 we complete the task by considering the issue of prosody synthesis, the job of generating phonetic or acoustic information from the prosodic form. As we shall see, the degree to which prosody needs to be specified depends very much on the synthesis technique employed... 

We need to reiterate some points about the augmentative function of prosody and how this affects studies of prosodic phenomena. While it is widely accepted that prosody does undergo more variation, or is subject to more speaker choice, than other parts of language, it is our belief that the degree to which this occurs is vastly underestimated in most studies. If we compare prosodic form with verbal form, we see that, while speakers can indeed vary the articulation of a word such as hat, they are very constrained in so doing, and too much deviation from the canonical pronunciation runs the risk of the word being decoded incorrectly. 

---