Datasets:
Languages:
Luxembourgish
License:
Lemswasabi
commited on
Commit
·
903d3e2
1
Parent(s):
7095a78
Create luxembourgish-asr-rtl-lu
Browse files- luxembourgish-asr-rtl-lu +100 -0
luxembourgish-asr-rtl-lu
ADDED
@@ -0,0 +1,100 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
#!/usr/bin/env python3
|
2 |
+
#
|
3 |
+
# Created by lemswasabi on 17/05/2022.
|
4 |
+
# Copyright © 2022 letzspeak. All rights reserved.
|
5 |
+
#
|
6 |
+
"""Luxembourgish ASR RTL.lu Dataset"""
|
7 |
+
|
8 |
+
|
9 |
+
import os
|
10 |
+
|
11 |
+
import datasets
|
12 |
+
|
13 |
+
from datasets.tasks import AutomaticSpeechRecognition
|
14 |
+
|
15 |
+
|
16 |
+
_DESCRIPTION = """\
|
17 |
+
luxembourgish-asr-rtl-lu dataset is a speech corpus for the under-resourced Luxembourgish language.
|
18 |
+
"""
|
19 |
+
|
20 |
+
_URLS = {
|
21 |
+
"rtl-benchmark": "https://drive.google.com/uc?id=1IiFV6TZHH1sOBL409VnmxCXSSyQkue0F&export=download&confirm=t",
|
22 |
+
}
|
23 |
+
|
24 |
+
class Tuudle(datasets.GeneratorBasedBuilder):
|
25 |
+
|
26 |
+
VERSION = datasets.Version("1.1.0")
|
27 |
+
|
28 |
+
BUILDER_CONFIGS = [
|
29 |
+
datasets.BuilderConfig(name="rtl-benchmark", version=VERSION, description="This part contains benchmark of samples collected from the RTL.lu domain"),
|
30 |
+
]
|
31 |
+
|
32 |
+
DEFAULT_CONFIG_NAME = "tuudle"
|
33 |
+
|
34 |
+
def _info(self):
|
35 |
+
|
36 |
+
features = datasets.Features(
|
37 |
+
{
|
38 |
+
"audio": datasets.Audio(sampling_rate=16_000),
|
39 |
+
"sentence": datasets.Value("string"),
|
40 |
+
}
|
41 |
+
)
|
42 |
+
|
43 |
+
return datasets.DatasetInfo(
|
44 |
+
description=_DESCRIPTION,
|
45 |
+
features=features,
|
46 |
+
supervised_keys=("audio", "sentence"),
|
47 |
+
task_templates=[AutomaticSpeechRecognition(audio_column="audio", transcription_column="sentence")],
|
48 |
+
)
|
49 |
+
|
50 |
+
def _split_generators(self, dl_manager):
|
51 |
+
|
52 |
+
urls = _URLS[self.config.name]
|
53 |
+
archive_path = dl_manager.download_and_extract(urls)
|
54 |
+
metadata_filepaths = {
|
55 |
+
split: os.path.join(archive_path, os.path.join(split, f"{split}.tsv"))
|
56 |
+
for split in ["train", "test", "dev"]
|
57 |
+
}
|
58 |
+
|
59 |
+
return [
|
60 |
+
datasets.SplitGenerator(
|
61 |
+
name=datasets.Split.TRAIN,
|
62 |
+
gen_kwargs={
|
63 |
+
"local_extracted_archive": archive_path,
|
64 |
+
"metadata_filepath": metadata_filepaths["train"],
|
65 |
+
"split": "train",
|
66 |
+
},
|
67 |
+
),
|
68 |
+
datasets.SplitGenerator(
|
69 |
+
name=datasets.Split.TEST,
|
70 |
+
gen_kwargs={
|
71 |
+
"local_extracted_archive": archive_path,
|
72 |
+
"metadata_filepath": metadata_filepaths["test"],
|
73 |
+
"split": "test",
|
74 |
+
},
|
75 |
+
),
|
76 |
+
datasets.SplitGenerator(
|
77 |
+
name=datasets.Split.VALIDATION,
|
78 |
+
gen_kwargs={
|
79 |
+
"local_extracted_archive": archive_path,
|
80 |
+
"metadata_filepath": metadata_filepaths["dev"],
|
81 |
+
"split": "dev",
|
82 |
+
},
|
83 |
+
),
|
84 |
+
]
|
85 |
+
|
86 |
+
def _generate_examples(self, local_extracted_archive, metadata_filepath, split):
|
87 |
+
|
88 |
+
path_to_clips = os.path.join(local_extracted_archive, split)
|
89 |
+
|
90 |
+
with open(metadata_filepath, encoding="utf-8") as f:
|
91 |
+
lines = f.readlines()
|
92 |
+
for key, line in enumerate(lines[1:]):
|
93 |
+
field_values = line.strip().split("\t")
|
94 |
+
if len(field_values) == 2:
|
95 |
+
audio_filename, sentence = field_values[0], field_values[1]
|
96 |
+
audio_path = os.path.join(path_to_clips, audio_filename)
|
97 |
+
yield key, {
|
98 |
+
"audio": {"path": audio_path, "bytes": open(audio_path, "rb").read()},
|
99 |
+
"sentence": sentence,
|
100 |
+
}
|