(Page 4 of 5 in this chapter)
1.3 Preparing Text
- Applications speak text by passing text buffers to the function ttsSpeak. NaturalText sends these buffers to the DSP firmware which converts them into speech. The DSP firmware speaks whatever it finds in the buffer, regardless of content. Applications can change text and influence the sound of output speech by using one or more of the following NaturalText features:
- English E-mail Preprocessor
- Embedded Control Characters and Phonemes
1.3.1 User Exception Dictionaries
- NaturalText pronounces text in a highly intelligible, natural sounding voice. However, you can directly control the way NaturalText pronounces particular words (for instance, non-standard words or abbreviations) by creating custom User Exception Dictionaries (UED). User Exception Dictionaries automatically scan text passed to ttsSpeak for targeted words, and replace these words with substitute text. UEDs can establish set pronunciations for the abbreviations, acronyms, and specialized terms used by different organizations. For instance, a UED can specify that the name "Hrbec" is pronounced "Herbeck", or that the abbreviation "ECO" is expanded to "engineering change order".
- User Exception Dictionaries accomplish the following:
- Expand abbreviations and acronyms.
- Provide phonetic spellings for target words.
1.3.2 English E-mail Preprocessor
- E-mail message text is different from other types of text in several ways. E-mail messages contain lengthy headers, and specialized jargon not found in other documents. For English e-mail messages, the NaturalText API includes two functions (ttsPreprocess and ttsPreprocessFile) for filtering message headers and interpreting the specialized words, symbols and abbreviations.
- The E-mail Preprocessor performs the following functions:
- Expands abbreviations to pronounce full words or phrases.
- Pronounces common surnames, place names, phone numbers, and addresses
correctly.
- Interprets multiple date and time formats.
- Understands and interprets specialized punctuation (for instance, *asterisks*
in e-mail messages) to properly pronounce text.
- Assesses word context when pronouncing ambiguous punctuation or
abbreviations.
- Recognizes and compensates for missing punctuation.
- Expands emoticons (e-mail specific character sequences used to express
emotion). For example: ;-) is expanded to "wink".
1.3.3 Embedded Control Characters
- NaturalText firmware recognizes special control characters. Characteristics of the spoken text can be altered by embedding control characters in the text buffer before calling the function ttsSpeak. Control characters can be used to do the following:
- Control attributes such as emphasis, pitch, and speed.
- Turn phoneme interpretation on or off.
- See the NaturalText Text-to-Speech Converter Reference Manual for more information about control characters and phoneme input.
(Page 4 of 5 in this chapter)
tech_support@nmss.com
Copyright © 1997, Natural MicroSystems, Inc. All rights
reserved.