This paper presents some relevant aspects of the pronunciation of proper names and common lexica in European Portuguese. It presents statistical data concerning the occurrence and distribution of graphemes and phonemes for the two corpora and distinguishes between different subclasses found in proper names, namely first and last names, toponyms and acronyms. It also compares the performance of a rule-based method for letter-to-phone conversion for the two corpora with two self-learning methods, one based on a multi-layered neural network and another based on table look-up.