Languages that exist only in the spoken form, such as those of many of the Native American Indian tribes, require the use of audio recording campaign such as a digital audio recorder or a tape recorder, augmented by explanatory text written in other languages, for preservation. Recently, efforts have been made to create symbol sets and grammars for such languages to avoid further losses of historically valuable information. Conversely, there are numerous extinct languages which persist only in their printed form. In some cases, experts can only guess at what their spoken form really sounded like. To prevent further losses of this nature, modern linguists have developed supersets of symbols for the purpose of correctly recording the authentic sounds of known languages.
While the superset symbols, which normally encode phonemes and their allophones, could record continuous discussion with nearly flawless accuracy, it would be tedious and unnecessary to record huge volumes of speech in this manner. That said there is a practical application for it, fine tuning of the programming for text-to-speech synthesizers to give computers a voice that sounds extremely human. Where a synthesizer stumbles on a given word, an entry can be added to an omission dictionary, wherein its definition consists of the word respelled in phonemes and allophones.

No comments: