yemaozi88
2004399179
novoapi_functions.py is adjusted to use convert_phoneset.py.
2019-04-22 00:59:53 +02:00
yemaozi88
b444b70af9
fame_phonetics.py and functions to make quests.hed to tie triphone are added.
2019-03-25 00:06:53 +01:00
yemaozi88
bf586fcde5
triphone training is added.
2019-03-23 21:52:48 +01:00
yemaozi88
fdd165ce6a
re-aligned mlf file include less files than original mlf file. Therefore the scp file should also be updated accordingly, when re-estimation is performed. this bug is fixed.
2019-03-08 23:13:08 +01:00
yemaozi88
fa81b70b27
monophone training is completed.
2019-03-07 22:16:50 +01:00
yemaozi88
41d4fa5ff9
sp is added to the model.
2019-03-05 00:11:38 +01:00
yemaozi88
b1b1942fa0
test on stimmen data is added.
2019-03-03 02:05:37 +01:00
yemaozi88
c185072d5b
label alignment using HVite is added.
2019-02-14 00:21:28 +01:00
yemaozi88
8f89f60538
dataset is made.
2019-02-08 14:10:32 +01:00
yemaozi88
f6e563ecd3
moved testing parts in htk_vs_kaldi into stimmen_test.py
2019-02-06 09:35:23 +01:00
yemaozi88
da0242b0e1
make sure all the phones in stimmen transcription can be treated correctly.
2019-02-06 00:00:14 +01:00
yemaozi88
ab3887c6ca
sp is added to the model.
2019-02-04 20:32:12 +01:00
yemaozi88
f6e7c8eefa
bug related encoding on label file is fixed.
2019-02-04 13:46:27 +01:00
yemaozi88
322a8a0079
label files are extracted. hcompv_scp is made.
2019-02-03 13:54:37 +01:00
yemaozi88
22cccfb61d
fix the bug there are characters in the lexicon which cannot be described in ascii.
2019-02-03 00:34:35 +01:00
yemaozi88
dc6b7b84b6
lexicon is made.
2019-01-29 21:52:11 +01:00
yemaozi88
8cda93de75
fame_asr phoneset is added including reduced version and htk compatible version.
2019-01-28 12:34:20 +01:00
yemaozi88
87abbbb95a
correspondence between lex.asr and lex.ipa is automatically obtained. header is added to the functions in fame_functions.py.
2019-01-27 23:52:33 +01:00
yemaozi88
813f013d7a
phonset is given as fame_phoneset.py. translation key is obtained based on the information.
2019-01-27 01:34:04 +01:00
a.kunikoshi
7844a56281
HTK related functions are moved to pyhtk project. fame acoustic models are made using fame_hmm.py. feature extraction is completed. A function is being made to get translation key from ipa to asr.
2019-01-24 09:38:28 +01:00
a.kunikoshi
04a862b2fd
Merge branch 'master' of https://git.webhosting.rug.nl/p280427/acoustic_model
2019-01-21 21:57:46 +01:00
yemaozi88
24ac56ac0e
to transfer working environment to McRoberts laptop.
2019-01-21 21:56:55 +01:00
yemaozi88
82a8e2302f
rozen-test is added.
2019-01-21 10:35:50 +01:00
yemaozi88
de5c9cecb9
The bug regarding novoapi for Python 3.6 is solved. The detail can be found in novoapi_for_python3x/readme.txt
2019-01-20 13:47:29 +01:00
yemaozi88
8efb091715
confusion matrix is output.
2019-01-15 11:30:49 +01:00
yemaozi88
05e8a671c1
forced alignment by novoapi is performed.
2019-01-13 23:36:02 +01:00
yemaozi88
6edde06a4f
check frequency of the pronunciation variants of each word.
2019-01-12 23:29:56 +01:00
yemaozi88
1622655542
functions are added to perform forced_alignment using novoapi. results can be written in novo70 or IPA.
2019-01-10 23:39:02 +01:00
yemaozi88
d6d5543d03
novoapi_functions is added to novo70 specific functions.
2019-01-07 23:27:02 +01:00
yemaozi88
d6e005b1cb
find pronunciation variants which all phones are in novo70.
2019-01-07 11:50:24 +01:00
yemaozi88
dd9e3d820b
started to check which words in stimmen transcription consists of only phones in novo70 phoneset.
2018-12-31 13:04:33 +01:00
yemaozi88
e5cf182a18
novo_api for python 3.x is added.
2018-12-30 23:47:55 +01:00
yemaozi88
a77ed9d4dd
resolve the conflicts.
2018-12-30 23:34:20 +01:00
yemaozi88
df046ffc26
forced_alignment_novo.py is removed (.gitignore). commit history is cleaned up.
2018-12-30 23:30:33 +01:00
yemaozi88
beff33fdf9
revert "load novo phoneset" to continue working.
2018-12-30 23:19:18 +01:00
yemaozi88
af785e51cf
commit to clean up
2018-12-30 23:14:26 +01:00
yemaozi88
9ec7c3c50b
commit to clean up.
2018-12-30 21:36:12 +01:00
yemaozi88
f04bb2a080
Revert "the script 'forced_alignment_novo.py' which is to run novo_api on Python 3.6 environment is added."
...
This reverts commit b87a81eb9d
2018-12-30 21:34:37 +01:00
yemaozi88
b771cbb7d6
just commit small change to revert.
2018-12-30 21:34:04 +01:00
yemaozi88
3500a8cdf0
load novo phoneset.
2018-12-30 21:27:20 +01:00
yemaozi88
b87a81eb9d
the script 'forced_alignment_novo.py' which is to run novo_api on Python 3.6 environment is added.
2018-12-30 21:26:42 +01:00
yemaozi88
0777735979
Code is cleaned up.
2018-09-16 23:33:31 +02:00
yemaozi88
ea30b5c503
The Stimmen excel file is loaded as Data Frame. Default values are given by defaultfiles.py.
2018-09-15 17:28:57 +02:00
yemaozi88
3a98e184fe
with bug-fixed xsampa->ipa conversion, FA is performed.
2018-09-02 12:17:21 +02:00
yemaozi88
df0e96c4f1
when convert_xsampa2ipa.conversion is used, '/' is recognized as escape sequence. to fix this bug, xsampa2ipa function is made.
2018-08-24 23:42:32 +02:00
yemaozi88
22b9ae966b
Forced alignment by Kaldi is added.
2018-08-20 22:50:53 +02:00
yemaozi88
d56ef7f075
FA result evaluation and xsampa to ipa conversion is updated.
2018-06-21 16:27:00 +02:00
yemaozi88
5fb05ddab2
xsampa to ipa conversion is added.
2018-06-18 00:15:50 +02:00
yemaozi88
a8dbb51d0c
to be sure.
2018-06-13 07:22:53 +09:00
yemaozi88
bbed340228
based on the recommendation from linguists, the total number of phones is reduced.
2018-04-25 09:07:46 +02:00
yemaozi88
a.kunikoshi