Skip to Main Content
In order to assess current efforts devoted to reading machine design, it is first necessary to develop a set of requirements for an ideal device. Direct translation aids are then seen to lack several of these desirable features, and more general, linguistically based techniques are then examined. Structural properties of English are found to be obtainable from the orthographic representation, and these abstract relations can then be used to infer structural correlates in the output speech waveform. Current knowledge is centered largely at the word level, but several correlates of higher order units have been studied and rules for their behavior have been implemented in working systems. Finally, direct assessment of speech synthesized by rule has shown that even currently available techniques can yield speech acceptable to blind users.