Title: Can a machine estimate the pronunciation difficulty of words?

Abstract:
The Graphon project is a Grapheme to Phoneme Converter for French. Supervised training methods are used to train the system to correctly assign a pronunciation to a word. The goal of my research is two-fold. The first goal is to improve the accuracy of the training methods by using larger datasets, weighted attributes, etc. These results could be applied to improve voice synthesis or to improve the content of Omnilex, a database of French words used by teachers and linguists. The final training methods will be used to assign a measure of difficulty to a given word. The second part goal is to determine, through studies with human subjects, if this measure of difficulty can be correlated to a human's difficulty in learning to read similar words.