Content such as news headlines that have a very wide range of vocabulary could still benefit from a locally stored audio symbol library, but unless the news was restricted to a very small domain, the symbols would likely have to be the phonetic units of the target language as rendered by a text to speech converter or voice synthesizer.