FYI, in http://unicode.org/cldr/trac/changeset/13540 I’ve changed the CLDR version of the Tamil-to-IPA transcription to mark diphthongs, for example மும்பை to [mumbaɪ̯] instead of [mumbaɪ]. You might want to consider doing the same.

Can you give me some advice?

Just wanted to understand how the 'universal_depot.far' file has been created?
Where are the .grm file for all those FSTs in the 'universal_depot.far' file?

Thank you,
Surendra

For creating a pronunciation of a word, we have to do two task (phoneme finding and splitting into syllable). I think spitting into syllable is a big deal. Expected format [1]. How can I do it programmatically?

[1] https://github.com/google/language-resources/blob/master/bn/data/lexicon.tsv

Are the total durations of the speech recordings in these datasets available anywhere? I'd love to know that without downloading them and figuring it out for myself, if possible.

Hi,
I want to create a question file for Tamil language. Can I use the "generate_hts_questions.py" given here, to produce a question file for Tamil language?

Hi,
I was trying to build a Bangla Festival voice according to this guideline. But I got an error in the training phase. The error messages are quite long, but the first one was this:
xargs: <path to festival>/build/festvox/src/ehmm/bin/FeatureExtraction: No such file or directory

Can anyone point out the reasons for this error and/or methods to solve this problem.

Thanks in advance.

Please let me know how can I build unicode.fst model? Many thanks.

@pasindud sir, how you trained the TTS Voice without merlin defined in (https://github.com/googlei18n/language-resources/blob/master/bn/festvox/README.md).

I'm not understanding how can I integrate merlin with festival. I also viewed (https://github.com/googlei18n/language-resources/blob/master/bn/merlin/README.md).

Thanks in advance.

Hi,

Thanks for this library. I am wondering if there is a function to check if the input is Zawgyi or Unicode encoding.

Hi,
no grammar file for *_depot.far, Is it convenient to release these files?

I see many .tsv files containing following text for example in Sundanese:
CARDINAL_MARKUP cardinal|integer:-1| mineus hiji

And also the .grm files have corresponding grammar definitions to parse these patterns in .tsv files.
Where are these patterns "cardinal|integer:-1|" useful in real world text normalisation?
Why not .grm files just export rules to parse real world examples like "-1" instead of markup?

I build a thai text to speech. I found issue. $ ./festival_utils/build_festvox_voice.sh ${WAV_FOLDER} th ${OUTPUT_VOICE_FOLDER} > log.txt

Log : https://gist.github.com/wannaphongcom/aa60b952f6913364b3f55f823ec297a8

Crowdsourced high-quality multi-speaker speech data sets, I see there are line_index_{gender}.tsv file.

In the tsv files, I can see the following data,

guf_03209_00443170675	જગતમાં કોઈ જીવ ન્નસ જંગમ છે અને કોઈ સ્થાવર.
guf_01414_00718082800	ગામમાં ભગવાન સ્વામિનારાયણનું મંદિર છે

Here, what are 03209 and 00443170675 for the first sample? Are they speaker and utterance id?

FYI, the Tamil-to-IPA rules are now in Unicode CLDR, with one small modification: http://unicode.org/cldr/trac/changeset/13170

Interestingly, changing this rule didn’t affect the unit tests.

I manually placed my utts created in festival into labels/cmu_us_my_voice/festival/utts.

When trying to make labels SIOD error is coming and tmp file is not created. The error comes as follows under each of the file name:

** SIOD ERROR: unbound variable : eof
gawk: fatal: cannot open file ./full_context_labels/labels//tmp' for reading (No such file or directory) gawk: fatal: cannot open file ./full_context_labels/labels//tmp' for reading (No such file or directory) **

Can someone please help me with this issue?

I'm trying to build phoneset for chinese language, still stuck on how to use apply_phonology.py. I've tried to use it and getting issues with regex

Here is what I'm getting when I try to build necessary files.

$ python apply_phonology.py cantonese_phonology.json cantonese/data/
Traceback (most recent call last):
  File "apply_phonology.py", line 774, in <module>
    main(sys.argv)
  File "apply_phonology.py", line 739, in main
    assert len(phoneset_paths) == 1
AssertionError

Hello. I want conversion of FestVox voices to Flite (for thai language #25 ). Can you recommend me?

"Columbia" and "Columbian" refer to locations in North America. The country is Colombia. The cited paper gets the spelling right, but this repo and the SLR website have misspelled it.

In my lexicon file, words are in the format as follows
("அ" nil (((a) 1)))
I'm actually trying to build Tamil voices using festival.
But when I'm trying to build prompts using " ./bin/do_build $PARALLEL build_prompts " , it gives errors as " ta_1566 PROMPTS
LEXICON: Word ஆண்டாள் (plus features) not found in lexicon " for all the sentences in txt.done.data file.
But all those words are in my lexicon.scm file. How to get over with this error?

Really nice work and strong baseline of text normalization!

I am looking for a tool to do english text normalization and find sparrowhawk to solve my problem. But only a en_toy in documentation in sparrowhawk repo.

Counld you please provide a better english grm resources for sparrowhawk text normalization?

Thanks a lot !

is three a any way using pretraned models at here ?
only command line usage example given
How to make predictions using python code ?

In [1], all files are written in .grm file. @pasindud is there any official documentation for learning grm. Because, I wanna contribute to Bn Text Normalization. Is it a compiler design language?

[1] https://github.com/google/language-resources/tree/master/bn/textnorm