Re: [Espeak-general] Catalan pronunciation issues

[Reece D. <ms...@go...>]

On 12 May 2010 19:51, Fergus Henderson <fe...@go...> wrote:
> Hi,
> Forwarding on a bug report from a user of translate.google.com.
> On Wed, May 12, 2010 at 17:14, Edu Jodas Samper wrote:
>> Hi,
>>
>> I noticed today that http://translate.google.com offers the option to
>> listen to the translations. I checked some of them for the Catalan
>> language and found a mistake. It concerns the words finishing in 'ia'.
>> Unlike Spanish, in Catalan 'ia' is not considered a diphthong when at
>> the end of a word and so, when there is no accent, the stress should
>> be placed at the 'i', since 'a' is a different syllable. For example:
>>
>> - The word 'secretaria' has the syllables: se-cre-ta-ri-a, and so the
>> stress should be in -ri-; currently the stress is placed in -ta-, as
>> in Spanish. Putting the stress in -ta- makes the word sound like
>> 'secretària', with an accent in tà, which is a valid word, but has a
>> different meaning
>> - More examples:
>>  gestoria: should be ges-to-ri-a, but it is pronounced like
>> ges-tò-ria, a word that does not exist AFAIK
>>  traïdoria: should be traï-do-ri-a, but it is pronounced like
>> traï-dò-ria, a word that does not exist AFAIK
>>
>> All should have the stress in -ri- unless the previous syllable is
>> accentuated.
>
> I had a quick glance at the sources.
> The file dictsource/ca_rules seems to be the place which defines the
> pronunciation rules for Catalan,
> but the fix is not obvious.  Any suggestions?

Since all of these end in 'ria', you could try something like this (in
the r or ri block):

   ria (_       'ri:@

ria tells eSpeak to match the text "ria"
(_ tells eSpeak that this is at the end of the word
' tells eSpeak that the next phonemes are to be stressed
ri: is the ri syllable -- a "ree" sound; this is stressed as a result of using '
@ is the a syllable -- it is not a properly vocalised "a" sound, but
is shortened (more like an "uh" sound)

This should make the "ria" at the end of the word always stressed.

For the accented characters removing the stress, you will need to add
these cases as well, e.g.:

   ària (_      3@ri:@

where 3@ is the "air" sound (I think -- look at a rule for pronouncing
à or similar for the actual result).

NOTE: You don't need the t in tà in the rule above, as it is only the
accent character you are interested in. In addition, you don't need
the primary stress marker ' as eSpeak will calculate stresses using a
set of rules.

HTH,
- Reece