[espeak-gaeilge] Re: Latest change to current version in Drop box

  • From: "Cearbhall O'Meadhra" <omeadhrac@xxxxxxx>
  • To: <espeak-gaeilge@xxxxxxxxxxxxx>
  • Date: Wed, 29 Aug 2012 22:26:53 +0100

                Dear All,

I have made a change to the set of files in the current version of eSpeak in
the drop box. We noticed that one step had been overlooked when first
setting up the Gaeilge synthesiser in Espeak. The Phoneme file needs to be
modified with two extra lines that point to the ph-Irish file and enable the
language label "ga" before the compile can be completed. A copy of the
modified Phoneme file is now in the current folder. The instructions in the
Documentation folder have been amended to reflect the missing step. The new
instruction is included in the eSpeak Compile Instructions located in the
Documentation folder.

If you are already running eSpeak Gaeilge on your system then you will have
already completed this step. 

All the best,


Cearbhall 

T: +353 (0)1 6364364 m: 08333 23487 E: omeadhrac@xxxxxxx
-----Original Message-----
From: espeak-gaeilge-bounce@xxxxxxxxxxxxx
[mailto:espeak-gaeilge-bounce@xxxxxxxxxxxxx] On Behalf Of RobbieS
Sent: 29 August 2012 20:08
To: espeak-gaeilge@xxxxxxxxxxxxx
Subject: [MAILSECURE-SPAM] [espeak-gaeilge] The issue of English vs. Irish
phonemes
Importance: Low

 

 

 

 

 

 

 

 

On Monday night, Cearbhall, Ronan and myself had a conference call on Skype
to discuss general strategy, and in particular, whether the current set of
phonemes being used were adequate or appropriate for the project.  A
combination of English phonemes are currently being used, and I contend that
the Irish language uses quite a different set of phonemes, for the most
part.  We decided, at least, to look to see if such a set of Irish phonemes
exist digitally, so that they might be used as our building blocks or jigsaw
pieces.  eSpeak may actually have them already: otherwise, TCD or UCD may
have them.  

 

I said that if such digitised material doesn?t exist already, I am prepared
to work with John Duddington (who is behind eSpeak), to make these phonemes.


 

I may be wrong, but I got the sense that Ronan wasn?t completely convinced
of the need to change the building-blocks being used.  I know that he has
put a lot of work into what is currently there, and it is not nice to think
that such effort would be in vain.  Ronan based his selection of English
phonemes on the wikipedia article ?Irish phonetics?, and I agreed to read
this article and give a brief review (or my opinion, at least).  I know that
if such a radical shift as I?m suggesting is to be made, I must explain the
case that shows it will be worth it.  Although I have an aesthetic
preference for Irish phonemes, it is their utility in terms of
intelligibility which is the primary factor which should be considered.  

 

I?ve read the Irish Phonology article in Wikipedia, using voiceover.  It
could be my screen-reader, but there are occasional errors, such as dheas =
jass, but for the most part, the article appears to be cosher, albeit prone
to generalisations.  The statement in the introduction that Irish consonants
tend to be velar would be hard to substantiate, and I?m not sure about the
labiodental v which is mentioned (would have thought that bilabial v was
more accurate).  

 

My reading of the article was hampered by my inability to read (and
decipher) the IPA symbols.  Appreciation of the technical terms used in the
article is a prerequisite to its utility.  I?m a bit rusty on these
(especially the terminology relating to vowel-sounds), but shouldn?t take
too long to brush up.  

 

There are many net-resources to help, not least Wikipedia itslef: 

http://en.wikipedia.org/wiki/List_of_consonants
<http://en.wikipedia.org/wiki/List_of_consonants> 

Thiere?s a lot of material here, and it isn?t easy going with
screen-readers.  I have begun a summary relevant to English and irish, but
in the meantime, here?s a brief map relating to English.  All the consonants
consist of one of the following variables, or a combination of them.  

 

1.  use or non-use of vocal-chords (voiced or voiceless).

2.  the sort of passage of air through:

2.1:   plosives involve the stopping of air, with a sudden release (e.g., b,
c, ch, d, g, k, p, q, t, x).

2.2:  fricatives involve a restricted but consistent passage of air: (e.g.,
f, h, s, sh, v).

2.3: Nasal (e.g., m and n).

2.4:  lateral (e.g., various L-sounds).

3.  Parts of the mouth used to make the sound:

3.1: lips and teeth (i.e., both lips = bilabial; lip on teenth =
labiodental, 

3.2:  part of the mouth touched by the tongue (gum-ridge or soft palate).

3.3:  part of the tongue used and (tip, middle, back).

3.4.  Shape of tongue: especially in subcontinental Indian languages, tongue
can be curved to make different sounds.  I don?t think it need concern us
too much.  

 

 

 

The comparisons made in the ?Irish Phonetics? wiki article,  between irish
and English phonemes are approximations for the benefit of an Anglophone
reader: the article is in English, after all.  The technical description of
the Irish consonants in the article are sound (excuse the pun), but are not
reflected in the phonemes currently being used in eSpeak-Ga.  The glide
sounds or vowels following on from these consonants should be an intrinsic
part of the same phoneme, rather than the current structure which has an
English consonant along-side a different vowel phoneme.  The current
structure, therefore, makes the building of good and intelligible Irish
sounds into a mammoth if not impossible task.  

 

By acquiring more suitable building blocks or jigsaw pieces (my metaphors
for phonemes), we make the task much easier, and even feasable at all.  Even
if we can?t get consonant and glide-sound in the same phoneme, we still need
to make sure that the difference between Irish and English consonant sounds
are reflected in our synthesizer.  For instnace, as noted by the article,
there are four types of ?l? sound in irish, and at least three types of
R-sound, and the difference between these is essential to the meanings of
words.  This diversity is a real help to the synthesizer-maker and the maker
alike.  

 

 

More suggestions:

1.  ?dó? (number and preposition) sound the same as ?deo? in espeak.

 

2.  ?Trí? in eSpeak-Ga sounds just like the number ?hree? in English.  This
could be alleviated by improving the slender-t.  eSpeak?s pronunciation of
?tigh?, for instance, is basically an English t-sound with a y-ound after it
to represent a slender t, and after this, the letter ?I? is pronounced, so
an English phonetic of the result is ?tyig?.  If the Irish slender-t was
more faithfully rendered, then we could be able to do away with the
y-element altogether.  In the Suggestions folder I stated that this slender
t would ideally be between the English ch-sound and the English t-sound: I
meant by this that the tongue-shape was between the two (a tongue-shape
similar to slender c, d, l, n, r and s in Irish.  That is, that the middle
of the tongue touches the palate and the tip of the tongue doesn?t touch
anything (making a sibilant sound).  

 

3.  The slender gh in Irish is made by vibrating the back of the tongue
against the palate (coived), and the same for ?ch? (except ?ch? is
unvoiced).  Currently, with eSpeak, they are both being sounded like the
English g or c.  

 

4.  The slender R needs to be improved.  Compare, for example, ?braon?
(meaning drop of liquid) and ?bréan? (meaning disgusting).  They sound
identical in eSpeak-Ga.  Last time, I mentioned the rs or rzh sound that can
be made by vibrating the tip of the tongue off the palate (in slender
cases), and keeping the tongue stationary in that position to make a sort of
whistling rs  sound (in the cass of broad double-r).  

 

5.  Espeak is doing funny things with ?naoi? (the number nine).  The broad-L
sound I mentioned in the previous email is similar in many ways to the irish
broad N when either are followed by aoi, in that they are both dental, with
the tip of the tongue (or just behind it) touching the upper-teeth.  The
only difference is that the N is nasal.  Comining away from this sound,
necessitates a broad glide-vowel with the tongue.  The Irish phonetics
article put it well when it describes a w-sound without the rounded lips.
But comparing the plosive initial consonant of ?caoi? with the nasal initial
consonant in ?naoi? was unnecessarily confusing.  How to put this broad-N
into voice-synthesis is the tricky bit, I suppose.  There is another broad-N
sound and two more slender-N sounds.  I suppose the important thing to do is
to at least make sure that all the slender Ns sound different to the broad
Ns.  

            

            

 

 

go n-éirí linn le chéile,
 
Robbie


===========================================================================The 
eSpeak-Gaeilge mailing list
Manage account or unsubscribe: //www.freelists.org/list/espeak-gaeilge
 Archives: //www.freelists.org/archives/espeak-gaeilge
Administrative contact: omeadhrac@xxxxxxx

=========================================================

Other related posts: