Donald Noble's Website

Everyone ¹ seems to be amazed by the new voice included as part of mac OS X Leopard, and called Alex. Yes, it is better than previous computerised voices, but it (he?) still has a long way to go.

A lot is being made of the “natural breathing and intonation”, which is good. But I found, when using it to read aloud web pages, that it put. thespaces inall the wrong. places. This actually made it more difficult to understand some paragraphs. Perhaps, if people punctuated text properly, this would be less of an issue. However, some people have a tendency—especially on the web—to write whole paragraphs in one long breath!

It is also supposed to be an “English” voice (see the feature list). However, it has picked up quite a number of Americanisms while being programmed in California. It miss-pronounced “towards” as “t-ords” missing out the all important w, “enquiry” as “n-KWIR-ee” (with the emphasis in completely the wrong place), and others that I can’t remember. It probably sounds vaguely “English” to Americans, however, as a Scot, it sounds nothing like those foreign people south of the border :p.

I think the last straw was trying to get it to read out the text from Stephen Fry’s blog. The two mistakes above were in the first paragraph, and that was before it tried to pronounce any of Stephen’s more complex “words” such as “gismoidal” (OK, the only one google result for that word is on Stephen’s blog at the moment, so I’ll let it off).

That was probably not a good place to test the speech synthesis, as Stephen has an eloquent grasp of the English language, and hearing a computer attempt to pronounce the French, Latin, and other foreign language phrases interspersed through the text was jarring to say the least.

Perhaps if these were marked up as such in the HTML, it might fair better. But I note that it does not read the text for acronyms, so maybe not. I haven’t bothered to check though, and won’t hold my breath.

That is not to say that it isn’t good. It is a huge improvement on previous computerised voices. I think it was just hyped up a bit too much. In the same way that voice recognition has been ‘99.9% accurate’ for the last 10 years, but they are still making ‘huge improvements’ with each new version!

I am sure this is something that will be improved upon greatly in the future, but computers won’t be replacing people reading aloud audiobooks any time soon. And I will still have to read webpages, rather than just listen to the computer, for the sake of my sanity, if nothing else.

——
1. Well, a lot of influential people on the web, including John Gruber of Daring Fireball

This entry was posted on Sunday 11th November 2007 in Computers and tagged Computers, future, language, mac OS X. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a comment, or trackback from your own site.