gender voice variants
yury.tarasievich at gmail.com
Thu Dec 20 13:27:18 CET 2012
Reading what follows, please keep in mind that
I'm not too familiar with English computational
linguistic terminology. So,
1) You are talking about a text/speech
construction, which isn't what IETF registry is
2) Constructing the text, generally, is a
complicated problem, and it does not require
"just" a distinction like four use cases
provided by you, not in the first place, and not
only those, anyhow. E.g., even in the "narrow"
use case (antropomorphised UI) you are omitting
the use of singular/plural forms, un-personated
(безличные) sentences, polite addressing (which
actually "cancels" genera distinction in forms
of East Sl. languages). In fact, for every
language, and at least for every language
family, you'd have to build a separate module,
concerned not only with minor issue of what
gender it'd assume (if any!), but also with
conversion of semantics to wording, observing
the grammar etc. Even if it's just "picking the
UI text cards out of pack" is planned for such UI.
Have it your way, of course, but I strongly feel
this subtag solves nothing. :)
On 12/20/2012 03:00 PM, Milos Rancic wrote:
> We are not talking here about language recognition, but about audio
> UIs, which could, anthropomorphized, do the next: Talk as female, talk
> as male, talk to a female and talk to a male.
> All four variants include different, gender-specific speech. Take this
> sentence below ("You said something, but I thought something else")
> and its variants:
> * Tы что-то сказал, но я думал о чем-то другом.
> * Tы что-то сказал, но я думала о чем-то другом.
> * Tы что-то сказала, но я думал о чем-то другом.
> * Tы что-то сказала, но я думала о чем-то другом.
> We are talking here about the subtags which would cover such variations.
More information about the Ietf-languages