Overview

Text normalization for sound.speech. It turns a string into the token list the grapheme-to-phoneme stage reads:

  { word = "hello" [, emph = true ] }   a lowercase word: a-z and apostrophe
  { punct = "." | "," | "?" | "!" }     a phrase or sentence boundary

It spells out numbers (integers, decimals, negatives: "42" becomes "forty two"; a comma between digit groups of three is a thousands separator), reads a token mixing letters and digits run by run ("pad3" is "pad three"), expands a small table of abbreviations, and reads *word* as emphasis. .;:!? and em-dashes become boundaries; other punctuation is dropped.

Functions