Datasets:
Additional example files
Browse files- align_text_and_audio.py +169 -0
- requirements.txt +5 -0
- scrape_audio_from_librivox.py +38 -0
align_text_and_audio.py
ADDED
@@ -0,0 +1,169 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# -*- coding: utf-8 -*-
|
2 |
+
import aeneas.globalconstants as gc
|
3 |
+
import pandas as pd
|
4 |
+
import os
|
5 |
+
import argparse
|
6 |
+
|
7 |
+
from aeneas.executetask import ExecuteTask
|
8 |
+
from aeneas.language import Language
|
9 |
+
from aeneas.syncmap import SyncMapFormat
|
10 |
+
from aeneas.task import Task
|
11 |
+
from aeneas.task import TaskConfiguration
|
12 |
+
from aeneas.textfile import TextFileFormat
|
13 |
+
from pydub import AudioSegment
|
14 |
+
|
15 |
+
|
16 |
+
parser = argparse.ArgumentParser(description='A program to download all the chapters in a given librivox URL.')
|
17 |
+
parser.add_argument("--text_dir",
|
18 |
+
help='Directory containing the txt files to allign the audio to',
|
19 |
+
required=True,
|
20 |
+
type=str)
|
21 |
+
parser.add_argument("--audio_path",
|
22 |
+
help="path to the chapter which are to be aligned.",
|
23 |
+
required=True,
|
24 |
+
type=str)
|
25 |
+
parser.add_argument("--aeneas_path",
|
26 |
+
help="path to save the allignments from aeneas",
|
27 |
+
required=True,
|
28 |
+
type=str)
|
29 |
+
parser.add_argument("--en_audio_export_path",
|
30 |
+
help="path to save the english audio fragments",
|
31 |
+
required=True,
|
32 |
+
type=str)
|
33 |
+
parser.add_argument("--total_alignment_path",
|
34 |
+
help="path to save the english audio fragments",
|
35 |
+
required=True,
|
36 |
+
type=str)
|
37 |
+
parser.add_argument("--librivoxdeen_alignment",
|
38 |
+
help="path to TSV file provided by LibriVoxDeEn for this particular book",
|
39 |
+
required=True,
|
40 |
+
type=str)
|
41 |
+
parser.add_argument("--aeneas_head_max",
|
42 |
+
help="max value (s) of the head for the aeneas alignment, depending on the book to be alligned this has to be tuned",
|
43 |
+
default = 0,
|
44 |
+
required=False,
|
45 |
+
type=int)
|
46 |
+
parser.add_argument("--aeneas_tail_min",
|
47 |
+
help="min value (s) of the tail for the aeneas alignment, depending on the book to be alligned this has to be tuned",
|
48 |
+
default = 0,
|
49 |
+
required=False,
|
50 |
+
type=int)
|
51 |
+
args = parser.parse_args()
|
52 |
+
|
53 |
+
#variables
|
54 |
+
number_files = len(os.listdir(args.text_dir))
|
55 |
+
txt_files = os.listdir(args.text_dir)
|
56 |
+
txt_files.sort()
|
57 |
+
|
58 |
+
audio_files = os.listdir(args.audio_path)
|
59 |
+
audio_files.sort()
|
60 |
+
|
61 |
+
EN_AUDIO_EXPORT_NAME = '000{}-{}.wav' # path where to store the english fragments
|
62 |
+
EN_ALIGNED_CSV = '000{}-undine_map.csv' #name to store the aeneas mapping
|
63 |
+
TOTAL_ALIGNMENT_NAME = '000{}-undine_map_DeEn.csv' #path to store the english and german mapping
|
64 |
+
|
65 |
+
GE_AUDIO_NAME_TEMPLATE = '000{}-undine_{}.flac' #name of german audio file
|
66 |
+
#aeneas arguments
|
67 |
+
tsv_audio_template = "000{}" #template of audio files in tsv file
|
68 |
+
|
69 |
+
#count the number of total missing files unaligned
|
70 |
+
total_missing = 0
|
71 |
+
total = 0
|
72 |
+
|
73 |
+
|
74 |
+
if not os.path.exists(args.aeneas_path):
|
75 |
+
os.makedirs(args.aeneas_path)
|
76 |
+
print('made new directory at:',args.aeneas_path)
|
77 |
+
if not os.path.exists(args.en_audio_export_path):
|
78 |
+
os.makedirs(args.en_audio_export_path)
|
79 |
+
print('made new directory at:',args.en_audio_export_path)
|
80 |
+
if not os.path.exists(args.total_alignment_path):
|
81 |
+
os.makedirs(args.total_alignment_path)
|
82 |
+
print('made new directory at:',args.total_alignment_path)
|
83 |
+
|
84 |
+
|
85 |
+
|
86 |
+
for chap in range(1 ,number_files+1):
|
87 |
+
|
88 |
+
abs_path = args.text_dir+txt_files[chap-1]
|
89 |
+
str_chap = str(chap)
|
90 |
+
str_chap = str(chap).zfill(2)
|
91 |
+
|
92 |
+
print('start alignent for chap: {}'.format(chap))
|
93 |
+
|
94 |
+
# create Task object
|
95 |
+
config = TaskConfiguration()
|
96 |
+
config[gc.PPN_TASK_LANGUAGE] = Language.ENG
|
97 |
+
config[gc.PPN_TASK_IS_TEXT_FILE_FORMAT] = TextFileFormat.PLAIN
|
98 |
+
config[gc.PPN_TASK_OS_FILE_FORMAT] = SyncMapFormat.CSV
|
99 |
+
config[gc.PPN_TASK_OS_FILE_NAME] = EN_ALIGNED_CSV.format(str_chap)
|
100 |
+
config[gc.PPN_TASK_IS_AUDIO_FILE_DETECT_HEAD_MAX] = args.aeneas_head_max
|
101 |
+
config[gc.PPN_TASK_IS_AUDIO_FILE_DETECT_TAIL_MIN] = args.aeneas_tail_min
|
102 |
+
config[gc.PPN_TASK_OS_FILE_HEAD_TAIL_FORMAT] = 'hidden'
|
103 |
+
|
104 |
+
task = Task()
|
105 |
+
task.configuration = config
|
106 |
+
task.text_file_path_absolute = abs_path
|
107 |
+
task.audio_file_path_absolute = args.audio_path+'/'+audio_files[chap-1]
|
108 |
+
task.sync_map_file_path = EN_ALIGNED_CSV.format(str_chap)
|
109 |
+
|
110 |
+
# process Task
|
111 |
+
ExecuteTask(task).execute()
|
112 |
+
|
113 |
+
task.output_sync_map_file(os.getcwd()+args.aeneas_path[1:])
|
114 |
+
|
115 |
+
# print(task.sync_map)
|
116 |
+
print('alignent done for chap: {}'.format(chap))
|
117 |
+
|
118 |
+
# -------- read csv and cut audio -----------
|
119 |
+
print('cutting audio for chap: {}'.format(chap))
|
120 |
+
|
121 |
+
en_alignment = pd.read_csv(args.aeneas_path+EN_ALIGNED_CSV.format(str_chap), header = None)
|
122 |
+
|
123 |
+
original = AudioSegment.from_file(args.audio_path+'/'+audio_files[chap-1])
|
124 |
+
for Name, Start, End, Text in zip(en_alignment.iloc[:,0], en_alignment.iloc[:,1], en_alignment.iloc[:,2], en_alignment.iloc[:,3]):
|
125 |
+
extract = original[Start*1000:End*1000]
|
126 |
+
extract.export(args.en_audio_export_path+EN_AUDIO_EXPORT_NAME.format(str_chap,Name), format="wav")
|
127 |
+
|
128 |
+
print('done cutting audio for chap: {}'.format(chap))
|
129 |
+
# ------- create CSV file with concatenated data ------
|
130 |
+
print('create CSV for chap: {}'.format(chap))
|
131 |
+
|
132 |
+
new = pd.DataFrame(columns=['book', 'DE_audio', 'EN_audio', 'score', 'DE_transcript', 'EN_transcript'])
|
133 |
+
|
134 |
+
# open German TSV
|
135 |
+
ge_alignment = pd.read_table(args.librivoxdeen_alignment)
|
136 |
+
|
137 |
+
# all german audio names
|
138 |
+
# restrict german to only the current chapter
|
139 |
+
ge_alignment = ge_alignment[ge_alignment['audio'].str.contains(tsv_audio_template.format(str_chap))==True]
|
140 |
+
ge_audio_names = []
|
141 |
+
|
142 |
+
for i in range(ge_alignment.shape[0]+1):
|
143 |
+
ge_audio_names.append(GE_AUDIO_NAME_TEMPLATE.format(str_chap,i))
|
144 |
+
|
145 |
+
for name in ge_audio_names:
|
146 |
+
ge_index = ge_alignment[ge_alignment['audio']==name].index
|
147 |
+
if len(ge_index) > 0:
|
148 |
+
en_translation = ge_alignment['en_sentence'][ge_index[0]].replace('<MERGE> ', '').strip()
|
149 |
+
en_index = en_alignment[en_alignment.iloc[:, 3] == en_translation.replace('~', '')].index
|
150 |
+
if len(en_index) > 0:
|
151 |
+
ind = ge_index[0]
|
152 |
+
new = new.append({'book': ge_alignment['book'][ind],
|
153 |
+
'DE_audio': ge_alignment['audio'][ind],
|
154 |
+
'EN_audio':'000{}-'.format(str_chap)+en_alignment.iloc[en_index[0],0]+'.wav',
|
155 |
+
'score':ge_alignment['score'][ind],
|
156 |
+
'DE_transcript':ge_alignment['de_sentence'][ind],
|
157 |
+
'EN_transcript': ge_alignment['en_sentence'][ind]},
|
158 |
+
ignore_index=True)
|
159 |
+
new.to_csv(args.total_alignment_path+'/'+TOTAL_ALIGNMENT_NAME.format(str_chap),index=False)
|
160 |
+
|
161 |
+
print('files in the DeEn csv: ' + str(new.shape))
|
162 |
+
print('files in the EN alignment'+str(en_alignment.shape))
|
163 |
+
print('files in the original DE csv: ' + str(ge_alignment.shape))
|
164 |
+
missing = ge_alignment.shape[0]-new.shape[0]
|
165 |
+
total_missing += missing
|
166 |
+
print('number of unaligned audio files: '+str(missing))
|
167 |
+
print('done creating CSV for chap: {}'.format(chap))
|
168 |
+
total += ge_alignment.shape[0]
|
169 |
+
print('missing: '+str(total_missing)+' out of '+str(total)+' GE files')
|
requirements.txt
ADDED
@@ -0,0 +1,5 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
aeneas=1.7.3.0
|
2 |
+
pandas>=1.1.4
|
3 |
+
pydub=0.24.1
|
4 |
+
beautifulsoup4=4.9.3
|
5 |
+
requests=2.25.1
|
scrape_audio_from_librivox.py
ADDED
@@ -0,0 +1,38 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# -*- coding: utf-8 -*-
|
2 |
+
import os
|
3 |
+
import argparse
|
4 |
+
import requests
|
5 |
+
from bs4 import BeautifulSoup
|
6 |
+
|
7 |
+
parser = argparse.ArgumentParser(description='A program to download all the chapters in a given librivox URL.')
|
8 |
+
parser.add_argument("--save_dir",
|
9 |
+
default="./chapters",
|
10 |
+
help='directory to save the downloaded chapters. If the directory does not exist it will be made.',
|
11 |
+
required=False,
|
12 |
+
type=str)
|
13 |
+
|
14 |
+
parser.add_argument("--url",
|
15 |
+
help='URL to audiobook.',
|
16 |
+
required=True,
|
17 |
+
type=str)
|
18 |
+
|
19 |
+
args = parser.parse_args()
|
20 |
+
|
21 |
+
page = requests.get(args.url)
|
22 |
+
soup = BeautifulSoup(page.content, 'html.parser')
|
23 |
+
results = soup.find_all(class_='chapter-name')
|
24 |
+
hrefs = []
|
25 |
+
for c in results:
|
26 |
+
hrefs.append(c['href'])
|
27 |
+
print('found {} chapters to download'.format(len(hrefs)))
|
28 |
+
|
29 |
+
if not os.path.exists(args.save_dir):
|
30 |
+
os.makedirs(args.save_dir)
|
31 |
+
print('made new directory at:',args.save_dir)
|
32 |
+
|
33 |
+
for audio in hrefs:
|
34 |
+
audio_path = args.save_dir+'/'+audio.split('/')[-1][:-3]+'wav'
|
35 |
+
print('writing {} to:'.format(audio),audio_path)
|
36 |
+
file = requests.get(audio)
|
37 |
+
with open(audio_path, 'wb') as f:
|
38 |
+
f.write(file.content)
|