I have developed several products where the voice messages are recorded as full sentences or as individual words that are "glued together" to form the messages.

As long as the correct person is selected (preferably a female, but with a bit deeper voice) for the recording, the users (quite often elderly) thinks that the readability is very high. With good editing, most people will not think about the words being concatenated on-the-fly.

The good thing is that todays flash memory prices allows quite long recording lenghts without significant costs. When merged with GPRS technology, there is also the possibility to do over-the-air updates to the messages, to add more bus stops or whatever is eneded.