I think you mixed up input/ouput because in your example at the end, you
would like to get pronunciation of a given new word.
input is the left hand side and output is the pron.
If you are able to rework a little bit the right hand side of your data
(you need to stretch the phones one by one, based on unique ones)
then the tool you are looking for is this one:
https://github.com/sequitur-g2p/sequitur-g2p
Le 26/04/2017 à 04:57, doc a écrit :
Many thanks for your kind interest.
Basically I aim was to find if a training tool exists which can train
data with the following format:
abc=def
Where the right hand is the input and the left hand is the output.
I chose English to UK IPA as an example. I have around 80,000+ strings
of English to IPA. An example is given below:
ˈem=əm
ˈneath=niːθ
ˈshun=ʃʌn
ˈtwas=twɒz
ˈtwen=twiːn
ˈtwen-decks=ˈtwiːn-deks
ˈtwere=twɜːr
ˈtwill=twɪl
ˈtwixt=twɪkst
ˈtwould=twʊd
ˈun=ən
A=eiː
Aˈs=eɪz
A-bomb=ˈeɪ-bɒm
If I had an unknown word like say
superpose
the tool after training should be able to predict
suːpəˈpəʊz
or something like that. Stress does not matter
I wonder if someone in the MOSES team could come up with a small tool
which can be used for such kind of work. As a linguist, I can assure
you that such a tool would be immensely popular and render great
service to the community.
I need not add that I will download GIZApp and try. I am trying out
SciKit, but the results are not very encouraging.
Many thanks
On Wed, Apr 26, 2017 at 3:23 AM, Allen Smith <allen.w.smi...@gmail.com
<mailto:allen.w.smi...@gmail.com>> wrote:
Are you wanting to train something to align letters and sounds, or
to figure out the sounds given the letters? (As it happens, I've
been working on using GIZApp to do the former.)
-Allen (Allen W. Smith, Ph.D.)
On Tue, Apr 25, 2017 at 4:01 AM, doc <raymond.doc...@gmail.com
<mailto:raymond.doc...@gmail.com>> wrote:
Hello,
I am looking for a tool for training data using either
Statistical methods or even CNN/RNN.
Basically the tool would allow the user to train simple data
and then once trained, it could be deployed to predict unknown
data
As an example, I have around 80,000 words in English converted
to IPA [text aligned in CSV format] and would like to train
the tool using the data, to predict in the case of new words.
Using Moses is like using a surgeon's scalpel to saw wood
And since I work in a Windows environment installing Moses is
not very easy
In any case a large number of linguists like me would prefer
to have a tool with ease of use.I am sure that if such a tool
is made available, it will be one of the most popular tools.
Thanks in advance for any help.
_______________________________________________
Moses-support mailing list
Moses-support@mit.edu <mailto:Moses-support@mit.edu>
http://mailman.mit.edu/mailman/listinfo/moses-support
<http://mailman.mit.edu/mailman/listinfo/moses-support>
_______________________________________________
Moses-support mailing list
Moses-support@mit.edu
http://mailman.mit.edu/mailman/listinfo/moses-support
_______________________________________________
Moses-support mailing list
Moses-support@mit.edu
http://mailman.mit.edu/mailman/listinfo/moses-support