Date   
[espeak-ng:master] reported: Need help improving Latin language #github

espeak-ng@groups.io Integration <espeak-ng@...>
 

[espeak-ng:master] New Comment on Issue #303 Need help improving Latin language
By nesrad:

I read the following on the site of the creator of the mbrola-la1 voice, which confirms what rhdunn has said:

Un certain nombre de problèmes peuvent se présenter, car toutes les combinaisons de phonèmes ne sont pas prises en compte dans mbrola en latin.

He offers the following solution using regular expressions:

On peut les éviter en modifiant certains phonèmes de manière quasi imperceptible; le script suivant (à adapter selon les besoins) fait des modifications dans le fichier test.txt et met le résultat dans test2.txt:

sed "s/1//g; s/2//g; s/3//g; s/4//g; s/5//g; s/6//g; s/7//g; s/8//g; s/9//g; s/0//g; s/^.//g; s/^$/, /g; s/à/a/g; s/ / /g;" test.txt|sed ':s;N;s/\n/ /;bs' | sed "s/mihi/miji/g; s/s s/s-s/g; s/t t/th t/g; s/c c/c\, c/g; s/c q/c\, q/g; s/m m/n m/g; s/m M/n M/g; s/mm/nm/g; s/aui/avi/g; s/ds/ts/g; s/d s/t s/g; s/eue/eve/g; s/aua/ava/g; s/aue/ave/g; s/eua/eva/g; s/r r/r\, r/g; s/dc/tc/g; s/d c/t c/g; s/bt/pt/g; s/b t/p t/g; s/b s/p s/g; s/bs/ps/g; s/eie/eje/g; s/eiu/eju/g; s/uiu/uju/g; s/aiu/aju/g; s/oiu/oju/g; s/poe/pee/g; s/ii/iji/g; s/ihi/iji/g; s/coe/cae/g; s/igg/ig/g; s/y/i/g; s/aie/aje/g; s/\//g; s/i h/i /g; s/dc/tc/g; s/d c/t c/g; s/d q/t q/g; s/dq/tq/g; s/aio/ajo/g; s/df/tf/g; s/d f/t f/g; s/n n/m n/g; s/d t/tt/g; s/dt/tt/g; s/boe/bee/g; s/foe/fee/g; s/b b/ b/g; s/b B/ B/g; s/Phoe/Phee/g; s/phoe/phee/g; s/d p/t p/g; s/dp/tp/g; s/d d/t d/g; s/cq/q/g; s/roe/ree/g; s/Roe/Ree/g; s/goe/gee/g; s/Goe/Gee/g; s/b p/p p/g; s/bp/pp/g"|cat > test2.txt

[espeak-ng:master] reported: Need help improving Latin language #github

espeak-ng@groups.io Integration <espeak-ng@...>
 

[espeak-ng:master] New Comment on Issue #303 Need help improving Latin language
By nesrad:

atferebantur and atpropinquare work well, but obzides is not good.

In response to your question: "how good is the espeak Latin voice?", the native espeak voice represents well the restored classical pronunciation, so it's a good starting point, although difficult to understand.

The mb-la1 voice is the best Latin voice that exists, and espeak is the only program capable of using it. No commercial TTS has Latin, and Google's "Latin" voice is nothing more than its Italian voice, useless for Classical Latin. So it would really be worthwhile improving espeak's support for this mbrola voice. The best way to do this would be to systematically ensure that each espeak phoneme has an acceptable equivalent for use with the mbrola voice, because my trial and error has only identified a certain number of problems, but I am sure that others exist. For example, I just discovered another one with "sā", like causā and mensārum. The ā is completely dropped. A work-around is to use "saā" as in causaā, mensaārum. The second a must have a macron in words like mensārum in order to place the stress accent on the correct syllable.

Btw, why do adFerebantur obSides adSum adPropinquare work with uppercase letters?

Github push to espeak-ng:espeak-ng #github

espeak-ng@groups.io Integration <espeak-ng@...>
 

1 New Commit:

[espeak-ng:master] By Reece H. Dunn <msclrhd@...>:
3a2663030b1b: Update the version to 1.49.3-dev.

Modified: CHANGELOG.md
Modified: android/jni/include/config.h
Modified: configure.ac
Modified: src/windows/config.h
Modified: src/windows/installer/Product.wxs

Updates to Github #github

espeak-ng@groups.io Integration <espeak-ng@...>
 

1 New Commit:

[espeak-ng:master] By Reece H. Dunn <msclrhd@...>:
3ed34d3a74eb: eSpeak NG 1.49.2

Modified: CHANGELOG.md
Modified: README.md
Modified: android/jni/include/config.h
Modified: configure.ac
Modified: src/windows/config.h


[espeak-ng] rhdunn pushed new tag 1.49.2.


[espeak-ng] created new tag 1.49.2.

Github push to espeak-ng:espeak-ng #github

espeak-ng@groups.io Integration <espeak-ng@...>
 

1 New Commit:

[espeak-ng:master] By Reece H. Dunn <msclrhd@...>:
4b4da735587e: mbrowrap: fix building on Windows.

Modified: src/libespeak-ng/mbrowrap.c

Updates to Github #github

espeak-ng@groups.io Integration <espeak-ng@...>
 

[espeak-ng:master] New Comment on Issue #303 Need help improving Latin language
By nesrad:

atferebantur and atpropinquare work well, but obzides is not good.

The mb-la1 voice is the best Latin voice that exists. No commercial TTS has Latin, and Google's "Latin" voice is nothing more than its Italian voice, useless for Classical Latin. So it would really be worthwhile improving espeak's support for this mbrola voice. The best way to do this would be to systematically ensure that each espeak phoneme has an acceptable equivalent for use with the mbrola voice, because my trial and error has only identified a certain number of problems, but I am sure that others exist. For example, I just discovered another one with "sā" ending a word, like causā. The final ā is completely dropped.

In response to your question: "how good is the espeak Latin voice?", the native espeak voice represents well the restored classical pronunciation, so it's a good starting point.

Btw, why do adFerebantur obSides adSum adPropinquare work with uppercase letters?


[espeak-ng:master] New Comment on Issue #303 Need help improving Latin language
By nesrad:

atferebantur and atpropinquare work well, but obzides is not good.

In response to your question: "how good is the espeak Latin voice?", the native espeak voice represents well the restored classical pronunciation, so it's a good starting point, although difficult to understand.

The mb-la1 voice is the best Latin voice that exists. No commercial TTS has Latin, and Google's "Latin" voice is nothing more than its Italian voice, useless for Classical Latin. So it would really be worthwhile improving espeak's support for this mbrola voice. The best way to do this would be to systematically ensure that each espeak phoneme has an acceptable equivalent for use with the mbrola voice, because my trial and error has only identified a certain number of problems, but I am sure that others exist. For example, I just discovered another one with "sā" ending a word, like causā. The final ā is completely dropped.

Btw, why do adFerebantur obSides adSum adPropinquare work with uppercase letters?


[espeak-ng:master] New Comment on Issue #303 Need help improving Latin language
By nesrad:

atferebantur and atpropinquare work well, but obzides is not good.

In response to your question: "how good is the espeak Latin voice?", the native espeak voice represents well the restored classical pronunciation, so it's a good starting point, although difficult to understand.

The mb-la1 voice is the best Latin voice that exists. No commercial TTS has Latin, and Google's "Latin" voice is nothing more than its Italian voice, useless for Classical Latin. So it would really be worthwhile improving espeak's support for this mbrola voice. The best way to do this would be to systematically ensure that each espeak phoneme has an acceptable equivalent for use with the mbrola voice, because my trial and error has only identified a certain number of problems, but I am sure that others exist. For example, I just discovered another one with "sā" ending a word, like causā. The final ā is completely dropped. A work-around is to use "saa" as in "causaa".

Btw, why do adFerebantur obSides adSum adPropinquare work with uppercase letters?


[espeak-ng:master] New Comment on Issue #303 Need help improving Latin language
By nesrad:

atferebantur and atpropinquare work well, but obzides is not good.

In response to your question: "how good is the espeak Latin voice?", the native espeak voice represents well the restored classical pronunciation, so it's a good starting point, although difficult to understand.

The mb-la1 voice is the best Latin voice that exists. No commercial TTS has Latin, and Google's "Latin" voice is nothing more than its Italian voice, useless for Classical Latin. So it would really be worthwhile improving espeak's support for this mbrola voice. The best way to do this would be to systematically ensure that each espeak phoneme has an acceptable equivalent for use with the mbrola voice, because my trial and error has only identified a certain number of problems, but I am sure that others exist. For example, I just discovered another one with "sā", like causā and mensārum. The ā is completely dropped. A work-around is to use "saā" as in causaā, mensaārum. The second a must have a macron in words like mensārum in order to place the stress accent on the correct syllable.

Btw, why do adFerebantur obSides adSum adPropinquare work with uppercase letters?


[espeak-ng:master] New Issue Created by nesrad:
#303 Need help improving Latin language

I have identified some problems with the translation of phonemes from Latin texts:

EDIT These problems affect the mbrola-la1 voice.

  1. In several cases, a consonant following another consonant is dropped. For example:
  2. adferebantur: espeak doesn't pronounce the f.
  3. obsides: doesn't pronounce the s (ob_ides). Also, subsidium, subsequi...
  4. adsum: doesn't pronounce the s (ad_um)
  5. adpropinquare: doesn't pronounce the d (a_propinquare) (throws an error in a sentence: "Bellovacorum adpropinquare cognoverant")
  6. there are probably other instances that I have not encountered yet.

  7. The diphthong "oe" is dropped. For example, in the word coeperint, espeak drops the letters oe. If you're wondering what the dipthong oe sounds like, the word coeperint should be pronounced the way espeak says "coēperint" (with a macron on the e).

  8. The consonant i is not pronounced (but it works fine with alternate spellings where j is used for consonant i). Thus, espeak drops the i inthe word "deiectus" , but "dejectus" works fine. Most modern editions of Latin texts use i instead of j (but both are acceptable spellings).

  9. Using the default word spacing causes some funky behaviour, like dropping leading letters ("legatum misit" gives "legatum _isit"). This is avoided by setting the word spacing to "-g 1".

EDIT: 5. Espeak drops "ā" (with macron) immediately after the letter s. Thus: causā becomes caus_, mensārum becomes mens_rum.

I do not know how to fix these problems myself. Any help would be appreciated.

Updates to Github #github

espeak-ng@groups.io Integration <espeak-ng@...>
 

[espeak-ng:master] New Comment on Issue #303 Need help improving Latin language
By nesrad:

atferebantur and atpropinquare work well, but obzides is not good.

The mb-la1 voice is the best Latin voice that exists. No commercial TTS has Latin, and Google's "Latin" voice is nothing more than its Italian voice, useless for Classical Latin. So it would really be worthwhile improving espeak's support for this mbrola voice. The best way to do this would be to systematically ensure that each espeak phoneme has an acceptable equivalent for use with the mbrola voice, because my trial and error has only identified a certain number of problems, but I am sure that others exist. For example, I just discovered another one with "sā" ending a word, like causā. The final ā is completely dropped.


[espeak-ng:master] New Comment on Issue #303 Need help improving Latin language
By nesrad:

atferebantur and atpropinquare work well, but obzides is not good.

The mb-la1 voice is the best Latin voice that exists. No commercial TTS has Latin, and Google's "Latin" voice is nothing more than its Italian voice, useless for Classical Latin. So it would really be worthwhile improving espeak's support for this mbrola voice. The best way to do this would be to systematically ensure that each espeak phoneme has an acceptable equivalent for use with the mbrola voice, because my trial and error has only identified a certain number of problems, but I am sure that others exist. For example, I just discovered another one with "sā" ending a word, like causā. The final ā is completely dropped.

Btw, why do adFerebantur obSides adSum adPropinquare work with uppercase letters?

[espeak-ng:master] reported: Need help improving Latin language #github

espeak-ng@groups.io Integration <espeak-ng@...>
 

[espeak-ng:master] New Comment on Issue #303 Need help improving Latin language
By rhdunn:

The problem is that the mbrola voices don't support all phoneme pairs. Thus, you get things like:

$ src/espeak-ng -x -v mb-la1 obsides
 'ObsIdEs
mbrola: Warning: b-s unkown, replaced with _-_

This affects all the mbrola voices. In this case, replacing the affected 's' with 'z' works (voicing the 's' after the voiced 'b' consonant). For 'df' in adferebantur, devoicing the 'd' to 't' before the unvoiced 'f' works; this is because la1 does not have 'v', so the 'f' cannot be voiced. Similarly, you can use 'tp' as a replacement for 'dp'. Do those work for you?

For coeperint, espeak pronounces it as k'OIpERInt, and the mbrola voice does not have a k-OI diphone (it uses OE for the oe sound). The coēperint word is pronounced as kO'e:pERInt. The espeak voice has a rule for 'oe' using OI, which should be changed to Oe:. Additionally, the mb-la1 voice should be mapping Oe: to OE to make use of the mbrola phoneme for the oe sound.

For deiectus, espeak is using dEjj'EktUs -- note the 'jj'. For dejectus it is using dEj'EktUs. This is similar to the issue with "legatum misit". That is, the MBROLA la1 voice does not support the doubled consonants (j-j and m-m in these cases). It does support those doubled phonemes as j: and m:, which espeak is not currently supporting.

Both the espeak Latin dictionary and MBROLA la1 voice are classical latin, so that should not be a problem. How good is the espeak Latin voice? Specifically, do the lack of voicing/devoicing rules affect the voice?

The general issue of missing diphones I am intending to address in https://github.com/espeak-ng/espeak-ng/issues/28. That issue is more complicated to support. In the meantime, making the above fixes should improve the mb-la1 voice.

[espeak-ng:master] new issue: Need help improving Latin language #github

espeak-ng@groups.io Integration <espeak-ng@...>
 

[espeak-ng:master] New Issue Created by nesrad:
#303 Need help improving Latin language

I have identified some problems with the translation of phonemes from Latin texts:

EDIT These problems affect the mbrola-la1 voice.

  1. In several cases, a consonant following another consonant is dropped. For example:
  2. adferebantur: espeak doesn't pronounce the f.
  3. obsides: doesn't pronounce the s (ob_ides). Also, subsidium, subsequi...
  4. adsum: doesn't pronounce the s (ad_um)
  5. adpropinquare: doesn't pronounce the d (a_propinquare) (throws an error in a sentence: "Bellovacorum adpropinquare cognoverant")
  6. there are probably other instances that I have not encountered yet.

  7. The diphthong "oe" is dropped. For example, in the word coeperint, espeak drops the letters oe. If you're wondering what the dipthong oe sounds like, the word coeperint should be pronounced the way espeak says "coēperint" (with a macron on the e).

  8. The consonant i is not pronounced (but it works fine with alternate spellings where j is used for consonant i). Thus, espeak drops the i inthe word "deiectus" , but "dejectus" works fine. Most modern editions of Latin texts use i instead of j (but both are acceptable spellings).

  9. Using the default word spacing causes some funky behaviour, like dropping leading letters ("legatum misit" gives "legatum _isit"). This is avoided by setting the word spacing to "-g 1".

EDIT: 5. Espeak drops the final "ā" (with macron) immediately after the letter s. Thus: causā becomes caus, pausā becomes paus.

I do not know how to fix these problems myself. Any help would be appreciated.

[espeak-ng:master] new issue: Need help improving Latin language #github

espeak-ng@groups.io Integration <espeak-ng@...>
 

[espeak-ng:master] New Issue Created by nesrad:
#303 Need help improving Latin language

I have identified some problems with the translation of phonemes from Latin texts:

EDIT These problems affect the mbrola-la1 voice.

  1. In several cases, a consonant following another consonant is dropped. For example:
  2. adferebantur: espeak doesn't pronounce the f.
  3. obsides: doesn't pronounce the s (ob_ides). Also, subsidium, subsequi...
  4. adsum: doesn't pronounce the s (ad_um)
  5. adpropinquare: doesn't pronounce the d (a_propinquare) (throws an error in a sentence: "Bellovacorum adpropinquare cognoverant")
  6. there are probably other instances that I have not encountered yet.

  7. The diphthong "oe" is dropped. For example, in the word coeperint, espeak drops the letters oe. If you're wondering what the dipthong oe sounds like, the word coeperint should be pronounced the way espeak says "coēperint" (with a macron on the e).

  8. The consonant i is not pronounced (but it works fine with alternate spellings where j is used for consonant i). Thus, espeak drops the i inthe word "deiectus" , but "dejectus" works fine. Most modern editions of Latin texts use i instead of j (but both are acceptable spellings).

  9. Using the default word spacing causes some funky behaviour, like dropping leading letters ("legatum misit" gives "legatum _isit"). This is avoided by setting the word spacing to "-g 1".

I do not know how to fix these problems myself. Any help would be appreciated.

Updates to Github #github

espeak-ng@groups.io Integration <espeak-ng@...>
 

[espeak-ng:master] New Issue Created by nesrad:
#303 Need help improving Latin language

I have identified some problems with the translation of phonemes from Latin texts:

EDIT These problems affect the mbrola-la1 voice.

  1. In several cases, a consonant following another consonant is dropped. For example:
  2. adferebantur: espeak doesn't pronounce the f.
  3. obsides: doesn't pronounce the s (ob_ides). Also, subsidium, subsequi...
  4. adsum: doesn't pronounce the s (ad_um)
  5. adpropinquare: doesn't pronounce the d (a_propinquare) (throws an error in a sentence: "Bellovacorum adpropinquare cognoverant")
  6. there are probably other instances that I have not encountered yet.

  7. The diphthong "oe" is dropped. For example, in the word coeperint, espeak drops the letters oe. If you're wondering what the dipthong oe sounds like, the word coeperint should be pronounced the way espeak says "coēperint" (with a macron on the e), but by running together the "o" and "ē" rather than pronouncing them succesively.

  8. The consonant i is not pronounced (but it works fine with alternate spellings where j is used for consonant i). Thus, espeak drops the i inthe word "deiectus" , but "dejectus" works fine. Most modern editions of Latin texts use i instead of j (but both are acceptable spellings).

  9. Using the default word spacing causes some funky behaviour, like dropping leading letters ("legatum misit" gives "legatum _isit"). This is avoided by setting the word spacing to "-g 1".

I do not know how to fix these problems myself. Any help would be appreciated.


[espeak-ng:master] New Comment on Issue #303 Need help improving Latin language
By nesrad:

By experimenting, I found that by replacing the letters that are dropped by UPPERCASE letters, the letters are pronounced correctly: adFerebantur obSides adSum adPropinquare

This does not work for the oe diphthong (coeperint), nor does it work for the i consonant (deiectus).

Updates to Github #github

espeak-ng@groups.io Integration <espeak-ng@...>
 

[espeak-ng:master] New Comment on Issue #303 Need help improving Latin language
By valdisvi:

Cannot reproduce it on development version of eSpeakNG.


[espeak-ng:master] New Comment on Issue #303 Need help improving Latin language
By valdisvi:

Cannot reproduce it on development version of eSpeakNG. But I know for sure that eSpeakNG spells this tested text differently than medical doctors in Latvia (probably eSpeakNG spells it as medical doctors from UK).


[espeak-ng:master] New Comment on Issue #303 Need help improving Latin language
By valdisvi:

Cannot reproduce it on development version of eSpeakNG. But I know for sure that eSpeakNG spells tested text differently than medical doctors in Latvia (probably eSpeakNG spells it as medical doctors from UK).


[espeak-ng:master] New Comment on Issue #303 Need help improving Latin language
By nesrad:

Hi valdisvi, I neglected to mention that the problems were encountered with the mbrola la1 voice. I will modify the ticket in consequence.


[espeak-ng:master] New Comment on Issue #303 Need help improving Latin language
By nesrad:

Hi valdisvi, I neglected to mention that the problems were encountered with the mbrola la1 voice. It doesn't seem to affect the regular voice. I will modify the ticket in consequence.


[espeak-ng:master] New Issue Created by nesrad:
#303 Need help improving Latin language

I have identified some problems with the translation of phonemes from Latin texts:

EDIT These problems affect the mbrola-la1 voice.

  1. In several cases, a consonant following another consonant is dropped. For example:
  2. adferebantur: espeak doesn't pronounce the f.
  3. obsides: doesn't pronounce the s (ob_ides). Also, subsidium, subsequi...
  4. adsum: doesn't pronounce the s (ad_um)
  5. adpropinquare: doesn't pronounce the d (a_propinquare) (throws an error in a sentence: "Bellovacorum adpropinquare cognoverant")
  6. there are probably other instances that I have not encountered yet.

  7. The diphthong "oe" is dropped. For example, in the word coeperint, espeak drops the letters oe. If you're wondering what the dipthong oe sounds like, the word coeperint should be pronounced the way espeak says "coēperint" (with a macron on the e), but without the pause between "o" and "ē".

  8. The consonant i is not pronounced (but it works fine with alternate spellings where j is used for consonant i). Thus, espeak drops the i inthe word "deiectus" , but "dejectus" works fine. Most modern editions of Latin texts use i instead of j (but both are acceptable spellings).

  9. Using the default word spacing causes some funky behaviour, like dropping leading letters ("legatum misit" gives "legatum _isit"). This is avoided by setting the word spacing to "-g 1".

I do not know how to fix these problems myself. Any help would be appreciated.

Updates to Github #github

espeak-ng@groups.io Integration <espeak-ng@...>
 

[espeak-ng:master] New Issue Created by nesrad:
#303 Need help improving Latin language

I have identified some problems with the translation of phonemes from Latin texts:

  1. In several cases, a consonant following another consonant is dropped. For example:
  2. adferebantur: doesn't pronounce the f.
  3. obsides: doesn't pronounce the s (ob_ides). Also, subsidium, subsequi...
  4. adsum: doesn't pronounce the s (ad_um)
  5. adpropinquare: doesn't pronounce the d (a_propinquare) (throws an error in a sentence: Bellovacorum adpropinquare cognoverant") there are probably other instances that I have not encountered yet.

  6. The diphthong "oe" is dropped. For example, in the word coeperint, espeak drops the letters oe. If you're wondering what the dipthong oe sounds like, the word coeperint should be pronounced the way espeak says "coēperint" (with a macron on the e), but without the pause between "o" and "ē".

  7. The consonant i is not pronounced (but it works fine with alternate spellings where j is used for consonant i). Thus, espeak drops the i inthe word "deiectus" , but "dejectus" works fine. Most modern editions of Latin texts use i instead of j (but both are acceptable spellings).

I do not know how to fix these problems myself. Any help would be appreciated.


[espeak-ng:master] New Issue Created by nesrad:
#303 Need help improving Latin language

I have identified some problems with the translation of phonemes from Latin texts:

  1. In several cases, a consonant following another consonant is dropped. For example:
  2. adferebantur: doesn't pronounce the f.
  3. obsides: doesn't pronounce the s (ob_ides). Also, subsidium, subsequi...
  4. adsum: doesn't pronounce the s (ad_um)
  5. adpropinquare: doesn't pronounce the d (a_propinquare) (throws an error in a sentence: "Bellovacorum adpropinquare cognoverant")
  6. there are probably other instances that I have not encountered yet.

  7. The diphthong "oe" is dropped. For example, in the word coeperint, espeak drops the letters oe. If you're wondering what the dipthong oe sounds like, the word coeperint should be pronounced the way espeak says "coēperint" (with a macron on the e), but without the pause between "o" and "ē".

  8. The consonant i is not pronounced (but it works fine with alternate spellings where j is used for consonant i). Thus, espeak drops the i inthe word "deiectus" , but "dejectus" works fine. Most modern editions of Latin texts use i instead of j (but both are acceptable spellings).

I do not know how to fix these problems myself. Any help would be appreciated.


[espeak-ng:master] New Issue Created by nesrad:
#303 Need help improving Latin language

I have identified some problems with the translation of phonemes from Latin texts:

  1. In several cases, a consonant following another consonant is dropped. For example:
  2. adferebantur: espeak doesn't pronounce the f.
  3. obsides: doesn't pronounce the s (ob_ides). Also, subsidium, subsequi...
  4. adsum: doesn't pronounce the s (ad_um)
  5. adpropinquare: doesn't pronounce the d (a_propinquare) (throws an error in a sentence: "Bellovacorum adpropinquare cognoverant")
  6. there are probably other instances that I have not encountered yet.

  7. The diphthong "oe" is dropped. For example, in the word coeperint, espeak drops the letters oe. If you're wondering what the dipthong oe sounds like, the word coeperint should be pronounced the way espeak says "coēperint" (with a macron on the e), but without the pause between "o" and "ē".

  8. The consonant i is not pronounced (but it works fine with alternate spellings where j is used for consonant i). Thus, espeak drops the i inthe word "deiectus" , but "dejectus" works fine. Most modern editions of Latin texts use i instead of j (but both are acceptable spellings).

I do not know how to fix these problems myself. Any help would be appreciated.


[espeak-ng:master] New Issue Created by nesrad:
#303 Need help improving Latin language

I have identified some problems with the translation of phonemes from Latin texts:

  1. In several cases, a consonant following another consonant is dropped. For example:
  2. adferebantur: espeak doesn't pronounce the f.
  3. obsides: doesn't pronounce the s (ob_ides). Also, subsidium, subsequi...
  4. adsum: doesn't pronounce the s (ad_um)
  5. adpropinquare: doesn't pronounce the d (a_propinquare) (throws an error in a sentence: "Bellovacorum adpropinquare cognoverant")
  6. there are probably other instances that I have not encountered yet.

  7. The diphthong "oe" is dropped. For example, in the word coeperint, espeak drops the letters oe. If you're wondering what the dipthong oe sounds like, the word coeperint should be pronounced the way espeak says "coēperint" (with a macron on the e), but without the pause between "o" and "ē".

  8. The consonant i is not pronounced (but it works fine with alternate spellings where j is used for consonant i). Thus, espeak drops the i inthe word "deiectus" , but "dejectus" works fine. Most modern editions of Latin texts use i instead of j (but both are acceptable spellings).

  9. Using the default word spacing causes some funky behaviour, like dropping leading letters ("legatum misit" gives "legatum _isit"). This is avoided by setting the word spacing to "-g 1".

I do not know how to fix these problems myself. Any help would be appreciated.

espeak-ng for windows daily builds over appveyor?

Simon Eigeldinger
 

Hi Reece,

I don't know if i have asked this.
Have you planned to set up nappveyor to compile espeak-ng for windows?
appveyor doesn't just provide daily builds it also does the same as travis.
at least that is what i have heard and seen.
would be interesting.

Greetings,
Simon

---
Diese E-Mail wurde von Avast Antivirus-Software auf Viren geprüft.
https://www.avast.com/antivirus

Github push to espeak-ng:espeak-ng #github

espeak-ng@groups.io Integration <espeak-ng@...>
 

1 New Commit:

[espeak-ng:master] By Reece H. Dunn <msclrhd@...>:
df7e9a23e1f7: Ensure that flags in the syllables are initialised. Clang static analysis reports this as 'Assigned value is garbage or undefined'.

Modified: src/libespeak-ng/intonation.c

Github push to espeak-ng:espeak-ng #github

espeak-ng@groups.io Integration <espeak-ng@...>
 

3 New Commits:

[espeak-ng:master] By Reece H. Dunn <msclrhd@...>:
19b2bebc0215: Don't save an intonation if it was not found. Clang static analysis reports an 'Assigned value is garbage or undefined' error. This happens if the default_tune data has not been copied in the kTUNE case. Extending this further, it does not make sense to save the tune data if the tune was not found, or if the tune already exists.

Modified: src/libespeak-ng/compiledata.c


[espeak-ng:master] By Reece H. Dunn <msclrhd@...>:
4a882e21d386: Don't update env_x/y if no points were read. Clang static analysis reports an 'Assigned value is garbage or undefined' for this. If no points were read from the file, the `n_points` variable is 0 and `env_x[n_points-1]` performs an out-of-bounds access.

Modified: src/libespeak-ng/compiledata.c


[espeak-ng:master] By Reece H. Dunn <msclrhd@...>:
57108e1ed7fa: Add n_points guards in the LoadEnvelope2 loop. Clang static analysis reports these as 'Result of operation is garbage or undefined' errors.

Modified: src/libespeak-ng/compiledata.c

Github push to espeak-ng:espeak-ng #github

espeak-ng@groups.io Integration <espeak-ng@...>
 

5 New Commits:

[espeak-ng:master] By Reece H. Dunn <msclrhd@...>:
988db184eaf1: Prevent invalid access to word in LookupDictList. Clang static analysis reports this as several 'Result of operation is garbage or undefined' errors.

Modified: src/libespeak-ng/dictionary.c


[espeak-ng:master] By Reece H. Dunn <msclrhd@...>:
e6ea8fabea98: Initialise ending in RemoveEnding. Clang static analysis reports this as several 'Result of operation is garbage or undefined' errors, when `ending` has not been set due to no matching endings.

Modified: src/libespeak-ng/dictionary.c


[espeak-ng:master] By Reece H. Dunn <msclrhd@...>:
8a777385a8e9: Use wflags to access wtab->flags in LookupDict2. Clang static analysis reports a 'Dereference of null pointer' error when accessing wtab->flags. This is properly guarded against when setting the wflags variable, so use that variable instead.

Modified: src/libespeak-ng/dictionary.c


[espeak-ng:master] By Reece H. Dunn <msclrhd@...>:
ecdff298b058: last_letter in MatchRule is not used. This was identified by the clang static analyser.

Modified: src/libespeak-ng/dictionary.c


[espeak-ng:master] By Reece H. Dunn <msclrhd@...>:
a2f751044ccb: Remove unused letter assignment in MatchRule. This was identified by the clang static analyser. The letter variable is set in the various match_type switch cases, so does not need to be initialised in the start of the while loop.

Modified: src/libespeak-ng/dictionary.c

[espeak-ng:master] reported: issues with georgian voice #github

espeak-ng@groups.io Integration <espeak-ng@...>
 

[espeak-ng:master] New Comment on Issue #302 issues with georgian voice
By beqabeqa473:

hi, what can be done to fix this?

Re: hawaiian language and faroese languages?

Timothy
 



On that note, how would one go about changing the consonances used for a given language? I am no linguist or programmer, but am a speaker of conversational Vietnamese. Some of the consonances aren't as pronounced as they need to be, and some are just wrong-sounding -- as if a non-native speaker spoke them. If I remember correctly, it has to do with manipulation of waveform data, but I also recall that process is inaccessible for a user of a screen reader.

Other changes, such as intonation and pronunciation of various words, can be modified only if one learns IPA, right? Those rules would also be inadvisable to changed without some sort of linguistical background as well? After all, you need to know the proper structure and given rules of a language in order to accurately write rule patterns for them... I can't just say "this sounds wrong -- it sounds like `a', but should rather sound like `b'..."

Edit: Never mind, I found some answers on this page:
https://github.com/espeak-ng/espeak-ng/blob/master/docs/add_language.md

Timothy

----- Original Message -----
Sent: Wednesday, September 6, 2017 1:23 PM
Subject: Re: [espeak-ng] hawaiian language and faroese languages?

On Wed, Sep 6, 2017 at 07:52 am, Josh Kennedy wrote:

also are there any plans to add the Hawaiian language and Faroese languages to espeak-ng?

There are no explicit plans to add any language to espeak-ng. However, I will not reject patches to add support for any language. It would be helpful if the person adding the language is familiar with it, so they can check that the pronunciations are reasonable for that language. Various people have added several languages to espeak-ng, compared to those in espeak.

Kind regards,
Reece.

 Josh

 

-- 
sent with mozilla thunderbird

Re: sapi5 espeak-ng?

Reece H. Dunn
 

On Wed, Sep 6, 2017 at 07:51 am, Josh Kennedy wrote:

hi

is there a good installable version of espeak-ng sapi5 yet?

Not yet. Implementation work is being tracked at https://github.com/espeak-ng/espeak-ng/issues/7. I don't have an estimate of when this will be completed, as there are many other things to work on in espeak-ng and I only have so much time to work on them.

or do I still have to use the old sourceforge espeak?

Currently, yes. Something that may work is to install espeak, then copy espeak-ng to that folder (renaming libespeak-ng.dll to libespeak.dll). This is not guaranteed to work, but as the APIs are compatible, it may work.

Kind regards,
Reece.

 Josh

 

-- 
sent with mozilla thunderbird