Adaptive networks as a model for human speech development

Unrestricted English text can be converted to speech through the use of a lookup table, or through a parallel feedforward network of deterministic processing units. The authors reproduce the network structure used in NETtalk. Several experiments have been carried out to determine which characteristics of the network were responsible for which learning behavior, and how closely the map resembles that of human speech development. The network was also trained with different levels of speech complexity, and with foreign languages. The results are shown to be highly dependent on statistical characteristics of the input.<>