[lojban] Re: Text-to-speech

[Arnt Richard Johansen <lojban-out@lojban.org>]

On Wed, Jan 16, 2008 at 01:31:07AM +0000, Jonathan Duddington wrote:
> I've added a Lojban voice to the development version of eSpeak speech
> synthesizer for Linux and Windows:
>   http://espeak.sourceforge.net/test/latest.html
> 
> The spelling-to-pronunciation rules are simple and regular, so that
> shouldn't be a problem, but I need advice on prosody.
> 
> Lojban differs from other languages in its lack of punctuation.  I
> recognise ".i" as a sentence marker, so I can break up a paragraph into
> sentences.  But the sentences also need pauses and intonation within
> them in order to sound natural.  In English, I would recognise commas
> and other punctuation as breaks.  Also conjunction words such as "and".
> 
> What lojban words should I look for to break a sentence into clauses
> (or their equivalent)?
> 
> This is not a problem of meaning or intelligibility.  It just sounds
> unnatural to speak a long sentence without using pause and intonation
> to indicate its structure (and to draw breath).

Nice! As you may know, I have been working on a Lojban diphone synthesizer, but that project has not been going anywhere for the past year, due to lack of time and energy to do the grunt work.

Here are some suggestions:

* ALWAYS have an intonation boundary before .i. It's the sentence separator, after all. Oh, and make sure that you separate cmavo clusters first, since people are apt to write .ije, .iku'i, and an infinite variety of combinations as one.
* ni'o should have an intonation boundary in front of it, and possibly a long pause.
* The following, and possibly some more, should have minor intonation boundaries in front of them: (na) gi'A nai (where A is any of aeiou); noi, poi, no'u, po'u; goi; to
* The following, and possibly some more, should have minor intonation boundaries after: toi

> Another question is which words to emphasize.  In English eSpeak has a
> list of common function words which are unstressed ("is", "the", "my",
> "of" etc).  For Lojban, I could make all one-syllable words (or even
> all cmavo) unstressed, but that's probably inappropriate.  I note that
> the pronunciation rules say that stressed syllables are optional for
> cmavo.

Do not stress any single-syllable cmavo. Unless you can insert a pause after it. ("le NO bliprenu" becomes "le nobli prenu", but "le NO. bliprenu" is okay.) Possibly not even then, since it sounds strange.

People often stress the first syllable in "go'i".

-- 
Arnt Richard Johansen                                http://arj.nvg.org/
- Hvorfor snakker man engelsk p� Internet?
- Har du h�rt om minste felles nevner?


To unsubscribe from this list, send mail to lojban-list-request@lojban.org
with the subject unsubscribe, or go to http://www.lojban.org/lsg2/, or if
you're really stuck, send mail to secretary@lojban.org for help.