Verbalisation

The final step in handling non-natural-language text is to convert it into words and this process is often called verbalisation. If we take our decoded-date example, we see that we have values for the tiiree fields of day, month and year, and with these we can 

Next we define a function yearQ. A simple version of this, which covers all years between 1010 and 1999, is 

Here we start to see the first case of why verbalisation can be tricky since some people prefer this rule not to have the and so that the function shonld produce 

Gregorian dates have a day a month and optionally a year and the task of the decoder is to determine the values of each of these from the text. The same date can give rise to a number of different text encodings, for example  

Chapter 5. Text Decoding Finding the words from the text 

In general, verbalisation functions for cases like these are not too difficult to construct, all that is required is a careful and thorough evaluation of the possibilities. The main difficulty in these cases is that there is no one correct expansion from 38 we see at least 7 ways of saying the date, so which should we pick While sometimes the original text can be a guide (such that 10/12/67 would be verbalised as ten twelve sixty seven) we find that human readers often do not do this, but rather pick a verbalisation that they believe is appropriate. The problem for us as TTS system builders is that sometimes the verbalisation we use is not the one that the reader expects or prefers, and this can be considered an error. 

---

Although superficially similar to real fighting, play fighting is distinct from it, and there are recognised cues which can be used in telling these two behaviours apart. These have been discerned by observational studies, and children too can verbalise many of these cues. 

Boyle s criteria and rules for making his preferred distinctions between matters of fact and causes have the status of conventions. Causal talk is grounded in conventions which Boyle s reports exemplify, just as the construction of the matter of fact is conventional in nature (...). The ultimate justification of convention does not take the form of verbalised rules. Instead the justification of convention is the form of life the total pattern of activities which includes discursive practices. 

Simply put, a speaker has a open choice of how much prosody or how much verbal content to put in a message they will balance these against each other, to create a final signal that conveys the message they wish in a way that is balanced between efficiency and effectiveness. A writer on the other hand has only a very limited ability to use prosody and so has to rely on the verbal component more. Because of this, he is likely to beef-up the verbal component, so that it is less ambiguous and more explicit than the spoken equivalent. In addition, the writer will verbalise any emotion if that is required (that is, e q)licitly explain the emotion by using words). 

For non-natural language classes, perform a verbalisation process, which converts their underlying forms into a sequence of words which can be spoken. 

Once a token s class is found, it is decoded into its underlying form and then translated into natural language by a process of verbalisation... 

Additional rules are used to shift prominence in cases where the parses would produce two prominent syllables side by side. We also find that many types of non-natural language entities have characteristic patterns. For example, it is common place to find telephone numbers read with a particular stress pattern (and indeed phrasing). Producing the correct stress patterns for these is often fairly straight forward and the creation of these patterns can be integrated into the verbalisation process. 

The situation is more difiicult in cases where there is more than one correct answer. This arises in verbalisation, where for example there are many possible ways of converting 3 15 into words, for example three fifteen or A quarter past three. One way of handling this is to mark all the possible outputs for a test sentence. This can prove impractical, so an alternative is to... 

---