Pronunciation resources will never be complete, as new words and expressions emerge. This project should allow users of ASR and TTS software to individually propose solutions to their pet-problems in pronunciation and have it automagically be fixed the next day.

A solution would implement a front-end (java applet or other) to allow for the manual addition of entries to a pronunciation resource from a queue of words with unknown pronunciation or for a word given by the user.

The user could be helped in determining the right pronunciation by showing the relevant parts of a corresponding Wiktionary-entry, or the proposition of an automatic G2P (such as espeak's). Also, the user should have easy and intuitive means of entering the phoneme symbols.

Ideally, pronunciations for all forms of a word would be registered at the same time, which means, that parts of speech have to be determined (either manually or from context using a POS tagger) and corresponding entry masks be shown.