Neural networks in text-to-speech systems for the Greek language

This paper presents the results of an experimental study for the use of neural networks in text-to-speech systems for the Greek language. In particular, neural networks are employed in text-to-phonemes conversion, where they demonstrate excellent performance. It is shown that even a small network can capture a significant portion of the regularities in the Greek pronunciation as well as absorb many of the irregularities. The results of the study show that the text-to-phonemes phase in a complete text-to-speech system for the Greek language can be implemented in real time with simple neural network architectures. The performance, with regards to the correct recognition of phonemes taken from Greek text in electronic form, is very good to excellent.