On Fri, 12 Sep 2003 23:56:38 -0400, Merritt L. Perkins mlperkins3@juno.com gave utterance to the following:
Having Encyclopedia articles read by a human voice would make the files so large that it would be impractical.
Not if is done via the standard technique: The data is sent to your computer as words + markup exactly as per a web page, and a client on your computer renders it as speech. Unfortunately, there is currently only one true speech browser (which parses HTML and aural css*) available - Emacspeak for Linux. The rest are page readers, which more or less speak the text which is output by an HTML renderer (usually embedded MSIE). Many are unaware of structural markup, relying on punctuation, and can run together text which has a strong visual separation.
*aural css allows control of speech properties.
Note that Opera has announce that it is working with IBM on greater incorporation of speech technology in web browsing. I think its mainly the sort of stuff where you can interact with a computer via speech on a cellphone - say when ordering a pizza. The "form" has speech prompts and your answers are processed via speech recognition.