dataset is made.

moved testing parts in htk_vs_kaldi into stimmen_test.py
make sure all the phones in stimmen transcription can be treated correctly.
2019-02-08 14:10:32 +01:00 · 2019-02-06 09:35:23 +01:00 · 2019-02-06 00:00:14 +01:00
16 changed files with 325 additions and 195 deletions
--- a/.vs/acoustic_model/v15/.suo
+++ b/.vs/acoustic_model/v15/.suo
--- a/acoustic_model/pycache/defaultfiles.cpython-36.pyc
+++ b/acoustic_model/pycache/defaultfiles.cpython-36.pyc
--- a/acoustic_model/acoustic_model.pyproj
+++ b/acoustic_model/acoustic_model.pyproj
@ -51,6 +51,8 @@
    <Compile Include="fame_hmm.py" />
    <Compile Include="phoneset\fame_asr.py" />
    <Compile Include="phoneset\fame_ipa.py" />
+    <Compile Include="stimmen_functions.py" />
+    <Compile Include="stimmen_test.py" />
  </ItemGroup>
  <ItemGroup>
    <Content Include="config.ini" />
--- a/acoustic_model/check_novoapi.py
+++ b/acoustic_model/check_novoapi.py
@ -20,12 +20,12 @@ from forced_alignment import convert_phone_set
 #import acoustic_model_functions as am_func
 import convert_xsampa2ipa
 import novoapi_functions
+import stimmen_functions
 sys.path.append(default.accent_classification_dir)
 import output_confusion_matrix

 ## procedure 
 forced_alignment_novo70 = True
-balance_sample_numbers  = False


 ## ===== load novo phoneset =====
@ -98,36 +98,7 @@ def search_phone_ipa(x, phone_list):


 ## ===== load all transcriptions (df) =====
-df = pd.read_excel(stimmen_transcription_, 'original')
-
-# mapping from ipa to xsampa
-mapping = convert_xsampa2ipa.load_converter('xsampa', 'ipa', default.ipa_xsampa_converter_dir)
-#for xsampa, ipa in zip(df['X-SAMPA'], df['IPA']):
-#    ipa_converted = convert_xsampa2ipa.xsampa2ipa(mapping, xsampa)
-#    if not ipa_converted == ipa:
-#        print('{0}: {1} - {2}'.format(xsampa, ipa_converted, ipa))
-
-ipas     = []
-famehtks = []
-for xsampa in df['Self Xsampa']:
-    if not isinstance(xsampa, float): # 'NaN'
-        # typo?
-        xsampa = xsampa.replace('r2:z@rA:\\t', 'r2:z@rA:t')
-        xsampa = xsampa.replace(';', ':')
-
-        ipa = convert_xsampa2ipa.xsampa2ipa(mapping, xsampa)
-        ipa = ipa.replace('ː', ':')
-        ipa = ipa.replace(' ', '')
-        ipas.append(ipa)
-    else:
-        ipas.append('')
-        
-# extract interesting cols.
-df = pd.DataFrame({'filename': df['Filename'], 
-                    'word': df['Word'], 
-                    'xsampa': df['Self Xsampa'],
-                    'ipa': pd.Series(ipas)})
-
+df = stimmen_functions.load_transcriptions()
 word_list = [i for i in list(set(df['word'])) if not pd.isnull(i)]
 word_list = sorted(word_list)

@ -184,21 +155,6 @@ if forced_alignment_novo70:
 		# samples in which all pronunciations are written in novo70.
 		samples = df_.query("ipa in @pronunciation_ipa")

-
-		## ===== balance sample numbers =====
-		if balance_sample_numbers:
-			c = Counter(samples['ipa'])
-			sample_num_list = [c[key] for key in c.keys()]
-			sample_num		= np.min(sample_num_list)
-
-			samples_balanced = pd.DataFrame(index=[], columns=list(samples.keys()))
-			for key in c.keys():
-				samples_ = samples[samples['ipa'] == key]
-				samples_balanced = samples_balanced.append(samples_.sample(sample_num), ignore_index = True)
-
-			samples = samples_balanced
-
-
 		results = pd.DataFrame(index=[],
 			columns=['filename', 'word', 'xsampa', 'ipa', 'result_ipa', 'result_novo70', 'llh'])
 	
--- a/acoustic_model/defaultfiles.py
+++ b/acoustic_model/defaultfiles.py
@ -2,63 +2,40 @@ import os
 # add path of the parent directory
 #os.path.dirname(os.path.realpath(__file__))

-#cygwin_dir = r'C:\cygwin64\home\Aki\acoustic_model'
-
-#htk_dir = r'C:\Aki\htk_fame'
-htk_dir = r'c:\OneDrive\Research\rug\experiments\acoustic_model\fame\htk'
-
-
-#config_hvite = os.path.join(cygwin_dir, 'config', 'config.HVite')
-#mkhmmdefs_pl = os.path.join(cygwin_dir, 'src', 'acoustic_model', 'mkhmmdefs.pl')
-
-#dbLexicon	  = C:\\Users\\Aki\\source\\repos\\rug_VS\\forced_alignment\\config\\lexicon.accdb
-#scriptBarbara = C:\\Users\\Aki\\source\\repos\\rug_VS\\forced_alignment\\config\\pronvars_barbara.perl
-#exeG2P		  = C:\\Users\\Aki\\source\\repos\\rug_VS\\forced_alignment\\config\\string2phon.exe
-
-#[pyHTK]
-#configHVite   = C:\\Users\\Aki\\source\\repos\\rug_VS\\forced_alignment\\config\\config.HVite
-#filePhoneList = C:\\Users\\Aki\\source\\repos\\rug_VS\\forced_alignment\\config\\phonelist_barbara.txt
-#AcousticModel = C:\\Users\\Aki\\source\\repos\\rug_VS\\forced_alignment\\config\\hmmdefs_16-2_barbara.compo
-
-#dbLexicon	  = config['cLexicon']['dbLexicon']
-#scriptBarbara = config['cLexicon']['scriptBarbara']
-#exeG2P		  = config['cLexicon']['exeG2P']
-
-#configHVite	  = config['pyHTK']['configHVite']
-#filePhoneList = config['pyHTK']['filePhoneList']
-#AcousticModel = config['pyHTK']['AcousticModel']
-
+# repos
 repo_dir = r'C:\Users\Aki\source\repos'
 ipa_xsampa_converter_dir    = os.path.join(repo_dir, 'ipa-xsama-converter')
 forced_alignment_module_dir = os.path.join(repo_dir, 'forced_alignment')
 accent_classification_dir   = os.path.join(repo_dir, 'accent_classification', 'accent_classification')
 toolbox_dir					= os.path.join(repo_dir, 'toolbox')

-#htk_config_dir = r'c:\Users\A.Kunikoshi\source\repos\forced_alignment\forced_alignment\data\htk\preset_models\aki_dutch_2017'
-#config_hvite = os.path.join(htk_config_dir, 'config.HVite')
-#acoustic_model = os.path.join(htk_config_dir, 'hmmdefs.compo')
-#acoustic_model = r'c:\cygwin64\home\A.Kunikoshi\acoustic_model\model\barbara\hmm128-2\hmmdefs.compo'
-phonelist_txt = os.path.join(htk_dir, 'config', 'phonelist.txt')
-
 WSL_dir   = r'C:\OneDrive\WSL'
-#fame_dir        = os.path.join(WSL_dir, 'kaldi-trunk', 'egs', 'fame')
-fame_dir = r'c:\OneDrive\Research\rug\_data\FAME'
+novo_api_dir = os.path.join(WSL_dir, 'python-novo-api', 'novoapi')
+#novo_api_dir = r'c:\Python36-32\Lib\site-packages\novoapi'

-fame_s5_dir     = os.path.join(fame_dir, 's5')
-fame_corpus_dir = os.path.join(fame_dir, 'corpus')
-
-experiments_dir = r'c:\OneDrive\Research\rug\experiments'
+# working directories
+rug_dir = r'c:\OneDrive\Research\rug'
+experiments_dir = os.path.join(rug_dir, 'experiments')
+htk_dir		= os.path.join(experiments_dir, 'acoustic_model', 'fame', 'htk')
 stimmen_dir = os.path.join(experiments_dir, 'stimmen')
-stimmen_data_dir = os.path.join(stimmen_dir, 'data')
+
+# data
+fame_dir = os.path.join(rug_dir, '_data', 'FAME')
+#fame_dir = os.path.join(WSL_dir, 'kaldi-trunk', 'egs', 'fame')
 # 44.1 kHz
 #stimmen_wav_dir  = os.path.join(stimmen_dir, 'wav')
 # 16 kHz
 stimmen_wav_dir  = r'c:\OneDrive\WSL\kaldi-trunk\egs\fame\s5\corpus\stimmen'
-stimmen_result_novoapi_dir = os.path.join(stimmen_dir, 'result', 'novoapi')
-
-stimmen_transcription_xlsx = os.path.join(stimmen_data_dir, 'Frisian Variants Picture Task Stimmen.xlsx')
+stimmen_transcription_xlsx = os.path.join(stimmen_dir, 'data', 'Frisian Variants Picture Task Stimmen.xlsx')
 phonelist_friesian_txt     = os.path.join(experiments_dir, 'friesian', 'acoustic_model', 'config', 'phonelist_friesian.txt')
-
-novo_api_dir = os.path.join(WSL_dir, 'python-novo-api', 'novoapi')
-#novo_api_dir = r'c:\Python36-32\Lib\site-packages\novoapi'
 novo70_phoneset = os.path.join(novo_api_dir, 'asr', 'phoneset', 'nl', 'novo70.phoneset')
+
+
+
+#phonelist_txt = os.path.join(htk_dir, 'config', 'phonelist.txt')
+#fame_s5_dir     = os.path.join(fame_dir, 's5')
+#fame_corpus_dir = os.path.join(fame_dir, 'corpus')
+#stimmen_result_novoapi_dir = os.path.join(stimmen_dir, 'result', 'novoapi')
+# novoapi_functions
+
+
--- a/acoustic_model/fame_functions.py
+++ b/acoustic_model/fame_functions.py
@ -341,3 +341,24 @@ def fix_single_quote(lexicon_file):

 def word2htk(word):
 	return ''.join([fame_asr.translation_key_word2htk.get(i, i) for i in word])
+
+
+def ipa2asr(ipa):
+	curr_dir = os.path.dirname(os.path.abspath(__file__))
+	translation_key_ipa2asr = np.load(os.path.join(curr_dir, 'phoneset', 'fame_ipa2asr.npy')).item(0)
+
+	ipa_splitted = convert_phoneset.split_word(ipa, fame_ipa.multi_character_phones)
+	ipa_splitted = fame_ipa.phone_reduction(ipa_splitted)
+	asr_splitted = convert_phoneset.convert_phoneset(ipa_splitted, translation_key_ipa2asr)
+	return ''.join(asr_splitted)
+
+
+def ipa2htk(ipa):
+	curr_dir = os.path.dirname(os.path.abspath(__file__))
+	translation_key_ipa2asr = np.load(os.path.join(curr_dir, 'phoneset', 'fame_ipa2asr.npy')).item(0)
+	
+	ipa_splitted = convert_phoneset.split_word(ipa, fame_ipa.multi_character_phones)
+	ipa_splitted = fame_ipa.phone_reduction(ipa_splitted)
+	asr_splitted = convert_phoneset.convert_phoneset(ipa_splitted, translation_key_ipa2asr)
+	htk_splitted = convert_phoneset.convert_phoneset(asr_splitted, fame_asr.translation_key_asr2htk)
+	return ''.join(htk_splitted)
--- a/acoustic_model/fame_hmm.py
+++ b/acoustic_model/fame_hmm.py
@ -27,7 +27,7 @@ extract_features  = 0
 flat_start		  = 0
 train_model_without_sp = 0
 add_sp = 0
-train_model_with_sp    = 0
+train_model_with_sp    = 1



@ -321,7 +321,8 @@ if add_sp:
 ## ======================= train model with short pause =======================
 if train_model_with_sp:
 	print('==== train model with sp ====')
-	for niter in range(niter_max+1, niter_max*2+1):
+	#for niter in range(niter_max+1, niter_max*2+1):
+	for niter in range(20, 50):
 		timer_start = time.time()
 		hmm_n = 'iter' + str(niter)
 		hmm_n_pre = 'iter' + str(niter-1)
--- a/acoustic_model/fame_test.py
+++ b/acoustic_model/fame_test.py
@ -69,6 +69,10 @@ else:
 translation_key_ipa2asr['ə:'] = 'ə'
 translation_key_ipa2asr['r.'] = 'r'
 translation_key_ipa2asr['r:'] = 'r'
+# added for stimmen.
+translation_key_ipa2asr['ɪ:'] = 'ɪ:'
+translation_key_ipa2asr['y:'] = 'y'
+
 np.save(os.path.join('phoneset', 'fame_ipa2asr.npy'), translation_key_ipa2asr)


--- a/acoustic_model/htk_vs_kaldi.py
+++ b/acoustic_model/htk_vs_kaldi.py
@ -1,131 +1,120 @@
 import os
 os.chdir(r'C:\Users\Aki\source\repos\acoustic_model\acoustic_model')
-
 import sys
-import csv
-import subprocess
-from collections import Counter
-import re
+
+#import csv
+#import subprocess
+#from collections import Counter
+#import re
+import shutil
+import glob

 import numpy as np
 import pandas as pd
-import matplotlib.pyplot as plt
-from sklearn.metrics import confusion_matrix
+#import matplotlib.pyplot as plt
+#from sklearn.metrics import confusion_matrix

-import acoustic_model_functions as am_func
-import convert_xsampa2ipa
+#import acoustic_model_functions as am_func
+#import convert_xsampa2ipa
 import defaultfiles as default

-from forced_alignment import pyhtk
+#from forced_alignment import pyhtk
+#sys.path.append(default.forced_alignment_module_dir)
+#from forced_alignment import convert_phone_set
+#import acoustic_model_functions as am_func
+import convert_xsampa2ipa
+import stimmen_functions
+import fame_functions
+import convert_phoneset
+from phoneset import fame_ipa, fame_asr
+sys.path.append(default.toolbox_dir)
+import file_handling as fh
+from htk import pyhtk


 ## ======================= user define =======================
-excel_file = os.path.join(default.experiments_dir, 'stimmen', 'data', 'Frisian Variants Picture Task Stimmen.xlsx')
-data_dir   = os.path.join(default.experiments_dir, 'stimmen', 'data')
+#excel_file = os.path.join(default.experiments_dir, 'stimmen', 'data', 'Frisian Variants Picture Task Stimmen.xlsx')
+#data_dir   = os.path.join(default.experiments_dir, 'stimmen', 'data')

-wav_dir	= r'c:\OneDrive\WSL\kaldi-trunk\egs\fame\s5\corpus\stimmen' # 16k
+#wav_dir	= r'c:\OneDrive\WSL\kaldi-trunk\egs\fame\s5\corpus\stimmen' # 16k

-acoustic_model_dir = os.path.join(default.experiments_dir, 'friesian', 'acoustic_model', 'model')
-htk_dict_dir       = os.path.join(default.experiments_dir, 'stimmen', 'dic_short')
-fa_dir             = os.path.join(default.experiments_dir, 'stimmen', 'FA_44k')
-result_dir         = os.path.join(default.experiments_dir, 'stimmen', 'result')
+#acoustic_model_dir = os.path.join(default.experiments_dir, 'friesian', 'acoustic_model', 'model')
+#htk_dict_dir       = os.path.join(default.experiments_dir, 'stimmen', 'dic_short')
+#fa_dir             = os.path.join(default.experiments_dir, 'stimmen', 'FA_44k')
+#result_dir         = os.path.join(default.experiments_dir, 'stimmen', 'result')

-kaldi_data_dir = os.path.join(default.kaldi_dir, 'data', 'alignme') 
-kaldi_dict_dir = os.path.join(default.kaldi_dir, 'data', 'local', 'dict')
-lexicon_txt    = os.path.join(kaldi_dict_dir, 'lexicon.txt')
+#kaldi_data_dir = os.path.join(default.kaldi_dir, 'data', 'alignme') 
+#kaldi_dict_dir = os.path.join(default.kaldi_dir, 'data', 'local', 'dict')
+#lexicon_txt    = os.path.join(kaldi_dict_dir, 'lexicon.txt')

 #lex_asr	 = os.path.join(default.fame_dir, 'lexicon', 'lex.asr')
 #lex_asr_htk = os.path.join(default.fame_dir, 'lexicon', 'lex.asr_htk')

+## procedure
+#make_htk_dict_files = 0
+#do_forced_alignment_htk = 0
+#eval_forced_alignment_htk = 0
+#make_kaldi_data_files = 0
+#make_kaldi_lexicon_txt = 0
+#load_forced_alignment_kaldi = 1
+#eval_forced_alignment_kaldi = 1

-# procedure
-make_htk_dict_files = 0
-do_forced_alignment_htk = 0
-eval_forced_alignment_htk = 0
-make_kaldi_data_files = 0
-make_kaldi_lexicon_txt = 0
-load_forced_alignment_kaldi = 1
-eval_forced_alignment_kaldi = 1
+#sys.path.append(os.path.join(default.repo_dir, 'forced_alignment'))
+#from forced_alignment import convert_phone_set
+#from forced_alignment import pyhtk
+
+#sys.path.append(os.path.join(default.repo_dir, 'toolbox'))
+#from evaluation import plot_confusion_matrix
+
+config_dir = os.path.join(default.htk_dir, 'config')
+model_dir = os.path.join(default.htk_dir, 'model')
+lattice_file = os.path.join(config_dir, 'stimmen.ltc')
+#pyhtk.create_word_lattice_file(
+#	os.path.join(config_dir, 'stimmen.net'), 
+#	lattice_file)
+hvite_scp = os.path.join(default.htk_dir, 'tmp', 'stimmen_test.scp')


+## ======================= load test data ======================
+stimmen_test_dir = r'c:\OneDrive\Research\rug\_data\stimmen_test'

+df = stimmen_functions.load_transcriptions_clean(stimmen_test_dir)
+df = stimmen_functions.add_row_asr(df)
+df = stimmen_functions.add_row_htk(df)

-## ======================= add paths =======================
-sys.path.append(os.path.join(default.repo_dir, 'forced_alignment'))
-from forced_alignment import convert_phone_set
-from forced_alignment import pyhtk
+word_list = [i for i in list(set(df['word'])) if not pd.isnull(i)]
+word_list = sorted(word_list)

-sys.path.append(os.path.join(default.repo_dir, 'toolbox'))
-from evaluation import plot_confusion_matrix
+# pronunciation variants
+for word in word_list:
+	df_ = df[df['word']==word]
+	print('{0} has {1} variants'.format(word, len(np.unique(df_['htk'])))

+#fh.make_filelist(stimmen_test_dir, hvite_scp, file_type='wav')

-## ======================= convert phones ======================
-mapping = convert_xsampa2ipa.load_converter('xsampa', 'ipa', default.ipa_xsampa_converter_dir)
+#output = pyhtk.recognition(
+#	os.path.join(default.htk_dir, 'config', 'config.rec', 
+#	lattice_file, 
+#	os.path.join(model_dir, 'hmm1', 'iter13'),
+#	dictionary_file, 
+#	os.path.join(config_dir, 'phonelist.txt'), 
+#	hvite_scp)

-xls = pd.ExcelFile(excel_file)
+			#pyhtk.create_label_file(
+			#	row['word'], 
+			#	os.path.join(stimmen_test_dir, filename.replace('.wav', '.lab')))

-## check conversion
-#df = pd.read_excel(xls, 'frequency')
-#for xsampa, ipa in zip(df['X-SAMPA'], df['IPA']):
-#    #ipa_converted = convert_xsampa2ipa.conversion('xsampa', 'ipa', mapping, xsampa_)
-#    ipa_converted = convert_xsampa2ipa.xsampa2ipa(mapping, xsampa)
-#    if not ipa_converted == ipa:
-#        print('{0}: {1} - {2}'.format(xsampa, ipa_converted, ipa))
-
-
-## check phones included in FAME!
-# the phones used in the lexicon.
-#phonelist = am_func.get_phonelist(lex_asr)
-
-# the lines which include a specific phone.
-#lines = am_func.find_phone(lex_asr, 'x')
-
-
-# Filename, Word, Self Xsampa
-df = pd.read_excel(xls, 'original')
-    
-ipas     = []
-famehtks = []
-for xsampa in df['Self Xsampa']:
-    if not isinstance(xsampa, float): # 'NaN'
-        # typo?
-        xsampa = xsampa.replace('r2:z@rA:\\t', 'r2:z@rA:t')
-        xsampa = xsampa.replace(';', ':')
-
-        ipa = convert_xsampa2ipa.xsampa2ipa(mapping, xsampa)
-        ipa = ipa.replace('ː', ':')
-        ipa = ipa.replace(' ', '')
-        ipas.append(ipa)
-        famehtk = convert_phone_set.ipa2famehtk(ipa)
-        famehtks.append(famehtk)
-    else:
-        ipas.append('')
-        famehtks.append('')
-        
-# extract interesting cols.
-df = pd.DataFrame({'filename': df['Filename'], 
-                    'word': df['Word'], 
-                    'xsampa': df['Self Xsampa'],
-                    'ipa': pd.Series(ipas),
-                    'famehtk': pd.Series(famehtks)})
-# cleansing. 
-df = df[~df['famehtk'].isin(['/', ''])]
-
-word_list = np.unique(df['word'])
-
-
-## ======================= make dict files used for HTK. ======================
-if make_htk_dict_files:
-    output_type = 3
-    
-    for word in word_list:
-        htk_dict_file = htk_dict_dir + '\\' + word + '.dic'
-
-        # pronunciation variant of the target word.
-        pronvar_ = df['famehtk'][df['word'].str.match(word)]
+## ======================= make a HTK dic file ======================
+#if make_htk_dic_file:
+#    output_type = 3
+dictionary_txt = os.path.join(default.htk_dir, 'lexicon', 'stimmen.dic') 
+#for word in word_list:
+word = word_list[2]
+# pronunciation variant of the target word.
+pronunciations = df_test['asr'][df_test['word'].str.match(word)]

        # make dic file.
-        am_func.make_htk_dict(word, pronvar_, htk_dict_file, output_type)
+        #am_func.make_htk_dict(word, pronvar_, htk_dict_file, output_type)
        

 ## ======================= forced alignment using HTK =======================
--- a/acoustic_model/novoapi_forced_alignment.py
+++ b/acoustic_model/novoapi_forced_alignment.py
@ -52,7 +52,7 @@ p = argparse.ArgumentParser()
 #p.add_argument("--user", default=None)
 #p.add_argument("--password", default=None)
 p.add_argument("--user", default='martijn.wieling')
-p.add_argument("--password", default='fa0Thaic')
+p.add_argument("--password", default='xxxxxx')
 args = p.parse_args()

 #wav_file = 'c:\\OneDrive\\WSL\\test\\onetwothree.wav'
--- a/acoustic_model/novoapi_functions.py
+++ b/acoustic_model/novoapi_functions.py
@ -173,7 +173,7 @@ def forced_alignment(wav_file, word, pronunciation_ipa):
 	# username / password cannot be passed as artuments...
 	p = argparse.ArgumentParser()
 	p.add_argument("--user", default='martijn.wieling')
-	p.add_argument("--password", default='fa0Thaic')
+	p.add_argument("--password", default='xxxxxx')
 	args = p.parse_args()
 	
 	rec = session.Recognizer(grammar_version="1.0", lang="nl", snodeid=101, user=args.user, password=args.password, keepopen=True) # , modeldir=modeldir)
--- a/acoustic_model/phoneset/fame_asr.py
+++ b/acoustic_model/phoneset/fame_asr.py
@ -73,12 +73,14 @@ reduction_key = {
 # already removed beforehand in phoneset. Just to be sure.
 phones_to_be_removed = ['ú', 's:', 'ɔ̈:']

-phoneset_short = [reduction_key.get(i, i) for i in phoneset
+def phone_reduction(phones):
+	return [reduction_key.get(i, i) for i in phones
 				  if not i in phones_to_be_removed]
-phoneset_short = list(set(phoneset_short))
+phoneset_short = list(set(phone_reduction(phoneset)))
 phoneset_short.sort()


+
 ## translation_key to htk format (ascii).
 # phones which gives UnicodeEncodeError when phone.encode("ascii")
 # are replaced with other characters.
--- a/acoustic_model/phoneset/fame_ipa.py
+++ b/acoustic_model/phoneset/fame_ipa.py
@ -5,6 +5,7 @@ phoneset = [
 	'i̯',
 	'i̯ⁿ',
 	'y',
+	'y:', # not included in lex.ipa, but in stimmen.
 	'i',
 	'i.',
 	'iⁿ',
@ -13,7 +14,7 @@ phoneset = [
 	'ɪ',
 	'ɪⁿ',
 	'ɪ.',
-	#'ɪ:', # not included in lex.ipa
+	'ɪ:', # not included in lex.ipa, but in stimmen.
 	'ɪ:ⁿ',
 	'e',
 	'e:',
@ -100,6 +101,36 @@ phoneset = [
 	'l'
 	]

+## reduce the number of phones.
+# the phones which are used in stimmen transcription but not in FAME corpus.
+# replacements are based on the advice from Jelske Dijkstra on 2018/06/21.
+stimmen_replacement = {
+	'æ': 'ɛ', 
+	'ø': 'ö', # or 'ö:'
+	'ø:': 'ö:', # Aki added.
+	'œ': 'ɔ̈', # or 'ɔ̈:'
+	'œ:': 'ɔ̈:', # Aki added.
+	'ɐ': 'a', # or 'a:'
+	'ɐ:': 'a:', # Aki added.
+	'ɑ': 'a', # or 'a:'
+	'ɑ:': 'a:', # Aki added
+	'ɒ': 'ɔ', # or 'ɔ:'
+	'ɒ:': 'ɔ:', # Aki added.
+	'ɾ': 'r',
+	'ʁ': 'r',
+	'ʊ': 'u',
+	'χ': 'x',
+
+	# aki guessed.
+	'ʀ': 'r',
+	'ɹ': 'r',
+	'w': 'ö'
+	}
+phoneset.extend(list(stimmen_replacement.keys()))
+
+def phone_reduction(phones):
+	return [stimmen_replacement.get(i, i) for i in phones]
+
 	
 ## the list of multi character phones. 
 # for example, the length of 'i̯ⁿ' is 3, but in the codes it is treated as one letter.
--- a/acoustic_model/phoneset/fame_ipa2asr.npy
+++ b/acoustic_model/phoneset/fame_ipa2asr.npy
--- a/acoustic_model/stimmen_functions.py
+++ b/acoustic_model/stimmen_functions.py
@ -0,0 +1,83 @@
+import os
+os.chdir(r'C:\Users\Aki\source\repos\acoustic_model\acoustic_model')
+import glob
+
+import pandas as pd
+
+import convert_xsampa2ipa
+import defaultfiles as default
+import fame_functions
+
+
+def _load_transcriptions():
+	stimmen_transcription = pd.ExcelFile(default.stimmen_transcription_xlsx)
+	df = pd.read_excel(stimmen_transcription, 'original')
+
+	# mapping from ipa to xsampa
+	mapping = convert_xsampa2ipa.load_converter('xsampa', 'ipa', default.ipa_xsampa_converter_dir)
+	#for xsampa, ipa in zip(df['X-SAMPA'], df['IPA']):
+	#    ipa_converted = convert_xsampa2ipa.xsampa2ipa(mapping, xsampa)
+	#    if not ipa_converted == ipa:
+	#        print('{0}: {1} - {2}'.format(xsampa, ipa_converted, ipa))
+
+	ipas     = []
+	for xsampa in df['Self Xsampa']:
+		if not isinstance(xsampa, float): # 'NaN'
+			# typo?
+			xsampa = xsampa.replace('r2:z@rA:\\t', 'r2:z@rA:t').replace(';', ':')
+
+			ipa = convert_xsampa2ipa.xsampa2ipa(mapping, xsampa)
+			ipa = ipa.replace('ː', ':').replace(' ', '')
+			ipas.append(ipa)
+		else:
+			ipas.append('')
+
+	df_ = pd.DataFrame({'filename': df['Filename'], 
+						'word': df['Word'], 
+						'xsampa': df['Self Xsampa'],
+						'ipa': pd.Series(ipas)})
+	
+	# not valid inputs, but seperator.
+	df_ = df_[~df_['ipa'].str.contains('/')] 
+	return df_.dropna()
+
+
+def load_transcriptions():
+	""" in default.stimmen_transcription_xlsx 
+	rows of which wav files can be easily found"""
+	df = _load_transcriptions()
+	df_ = pd.DataFrame(index=[], columns=list(df.keys()))
+	for index, row in df.iterrows():
+		filename = row['filename']
+		if isinstance(filename, str):
+			wav_file = os.path.join(default.stimmen_wav_dir, filename)
+			if os.path.exists(wav_file):
+				df_ = df_.append(row, ignore_index=True)
+	return df_
+
+
+def load_transcriptions_clean(clean_wav_dir):
+	df = _load_transcriptions()
+	wav_file_list = glob.glob(os.path.join(clean_wav_dir, '*.wav'))
+	df_clean = pd.DataFrame(index=[], columns=list(df.keys()))
+	for wav_file in wav_file_list:
+		filename = os.path.basename(wav_file)
+		df_ = df[df['filename'].str.match(filename)]
+		df_clean = pd.concat([df_clean, df_])
+	return df_clean
+
+
+def add_row_htk(df):
+	""" df['htk'] is made from df['ipa'] and added. """
+	htk = []
+	for index, row in df.iterrows():
+		htk.append(fame_functions.ipa2htk(row['ipa']))
+	return df.assign(htk=htk)
+
+
+def add_row_asr(df):
+	""" df['asr'] is made from df['ipa'] and added. """
+	asr = []
+	for index, row in df.iterrows():
+		asr.append(fame_functions.ipa2asr(row['ipa']))
+	return df.assign(asr=asr)
--- a/acoustic_model/stimmen_test.py
+++ b/acoustic_model/stimmen_test.py
@ -0,0 +1,64 @@
+import os
+os.chdir(r'C:\Users\Aki\source\repos\acoustic_model\acoustic_model')
+import sys
+import shutil
+
+#import numpy as np
+import pandas as pd
+
+import defaultfiles as default
+import convert_xsampa2ipa
+import stimmen_functions
+import fame_functions
+import convert_phoneset
+from phoneset import fame_ipa, fame_asr
+sys.path.append(default.toolbox_dir)
+import file_handling as fh
+from htk import pyhtk
+
+
+## ======================= user define =======================
+
+
+
+## ======================= make test data ======================
+stimmen_test_dir = r'c:\OneDrive\Research\rug\_data\stimmen_test'
+
+## copy wav files which is in the stimmen data.
+df = stimmen_functions.load_transcriptions()
+#for index, row in df.iterrows():
+#	filename = row['filename']
+#	wav_file = os.path.join(default.stimmen_wav_dir, filename)
+#	shutil.copy(wav_file, os.path.join(stimmen_test_dir, filename))
+
+# after manually removed files which has too much noise and multiple words...
+# update the info.
+df_clean = stimmen_functions.load_transcriptions_clean(stimmen_test_dir)
+
+# count how many files are removed due to the quality.
+word_list = [i for i in list(set(df['word'])) if not pd.isnull(i)]
+word_list = sorted(word_list)
+for word in word_list:
+	df_ = df[df['word']==word]
+	df_clean_ = df_clean[df_clean['word']==word]
+	print('word {0} has {1} clean files among {2} files ({3:.2f} [%]).'.format(
+		word, len(df_clean_), len(df_), len(df_clean_)/len(df_)*100))
+
+
+## check phones included in stimmen but not in FAME!
+splitted_ipas = [' '.join(
+		convert_phoneset.split_word(ipa, fame_ipa.multi_character_phones)) 
+		for ipa in df['ipa']]
+stimmen_phones = set(' '.join(splitted_ipas))
+stimmen_phones = list(stimmen_phones)
+fame_phones = fame_ipa.phoneset
+stimmen_phones.sort()
+fame_phones.sort()
+print('phones which are used in stimmen transcription but not in FAME corpus are:\n{}'.format(
+	set(stimmen_phones) - set(fame_phones)
+	))
+for ipa in df['ipa']:
+	ipa_splitted = convert_phoneset.split_word(ipa, fame_ipa.multi_character_phones)
+	if ':' in ipa_splitted:
+		print(ipa_splitted)
+
Author	SHA1	Message	Date
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