Italian language: secondary stress


ChrisLM
 

Hi,
in italian language secondary stress is used only in some exceptions;
words with four/five or more syllables, combined words etcc...

Espeak-ng adds secondary stress automatically on each word, also it changes the unstressed words indicated in it_list in secondary stress.
Is there a way to delete automatic secondary stress?

Can we use secondary stress only if is indicated in italian dictionaries or rules?

Thanks.
--
Chris.


Valdis Vitolins
 

Chris,
you have to adjust voice settings in
https://github.com/espeak-ng/espeak-ng/blob/master/espeak-ng-data/voice
s/roa/it file

Look at https://github.com/espeak-ng/espeak-ng/blob/master/docs/voices.
md#stresslength

For example, for Latvian settings are following:
...
stressAmp 14 10 10 8 0 0 20 15
stressLength 180 180 180 160 0 0 230 188
...

Note that 3rd group of numbers (secondary stress) is similar to 2nd
group (diminished, i.e. in general last syllable). (And I think maybe
I'll set 3rd group with even lower values, as espeak-ng still sometimes
stresses 3rd syllable too much).

Valdis

Hi,
in italian language secondary stress is used only in some exceptions;
words with four/five or more syllables, combined words etcc...

Espeak-ng adds secondary stress automatically on each word, also it 
changes the unstressed words indicated in it_list  in secondary
stress.
Is there a way to delete automatic secondary stress?

Can we use secondary stress only if is indicated in italian
dictionaries 
or rules?

Thanks.


Reece H. Dunn
 

I don't know this area of espeak to say what it's cababilities are in this regard. As for espeak-ng, write up a proposal in the issues describing the requirements you have for your needs w.r.t. Italian so they are documented.

I think espeak is weighted toward English stress patterns.

Ideally, two things should happen:

1.  Document the current stress behaviour in espeak;

2.  Investigate the different stress, tone and intonation patterns found in different languages and annotations for those languages (e.g. IPA tone marks);

3.  Design a framework that captures the requirements to support the things identified in (2);

4.  Implement (3) in espeak-ng.

I am currently looking at the voice/language/accent split within espeak-ng, so don't have time right now to look into improving the stress/tone/intonation support, but if someone else wants to look into this they can do.


ChrisLM
 

Hi,
by stressLength and stressAmp we can set amplitude and lengh, but not intonation.
We can set with lower value length and amplitude for secondary stress, but the intonation is always as secondary stress.

I need more informations before file a issues, I think that secondary stress should be set in tr_languages.c.
Like a case fr (French language), in the statement stress_flags.
E.G. see in src/libespeak-ng/tr_languages.c line 797:
"""
tr->langopts.stress_flags = S_NO_AUTO_2 | S_FINAL_DIM; // don't use secondary stress
"""




Chris.

Valdis Vitolins, 21/01/2017 14:41:

Chris,
you have to adjust voice settings in
https://github.com/espeak-ng/espeak-ng/blob/master/espeak-ng-data/voice
s/roa/it file

Look at https://github.com/espeak-ng/espeak-ng/blob/master/docs/voices.
md#stresslength

For example, for Latvian settings are following:
...
stressAmp 14 10 10 8 0 0 20 15
stressLength 180 180 180 160 0 0 230 188
...

Note that 3rd group of numbers (secondary stress) is similar to 2nd
group (diminished, i.e. in general last syllable). (And I think maybe
I'll set 3rd group with even lower values, as espeak-ng still sometimes
stresses 3rd syllable too much).