Datasets:
metadata
language:
- en
- zh
license: cc-by-sa-4.0
size_categories:
- 10K<n<100K
dataset_info:
- config_name: 30s
features:
- name: id
dtype: string
- name: audio
dtype:
audio:
sampling_rate: 16000
- name: transcription
dtype: string
- name: condition_on_prev
dtype: string
splits:
- name: train
num_bytes: 1117823922.576
num_examples: 1308
- name: validation
num_bytes: 115050795
num_examples: 135
- name: test
num_bytes: 117961106
num_examples: 138
download_size: 1207395166
dataset_size: 1350835823.576
- config_name: default
features:
- name: id
dtype: string
- name: path
dtype: string
- name: audio
dtype:
audio:
sampling_rate: 16000
- name: transcription
dtype: string
- name: duration
dtype: float32
- name: language
dtype: string
- name: original_speaker_id
dtype: int64
- name: session_id
dtype: int64
- name: topic
dtype: string
splits:
- name: train
num_bytes: 1014558975.36
num_examples: 9869
- name: test
num_bytes: 106170264.135
num_examples: 1315
- name: validation
num_bytes: 106771606.91
num_examples: 1130
download_size: 1223500329
dataset_size: 1227500846.4050002
configs:
- config_name: 30s
data_files:
- split: train
path: 30s/train-*
- split: validation
path: 30s/validation-*
- split: test
path: 30s/test-*
- config_name: default
data_files:
- split: train
path: data/train-*
- split: test
path: data/test-*
- split: validation
path: data/validation-*
Dataset Card for Dataset Name
This dataset is derived from CAiRE/ASCEND. More information is available at https://huggingface.co/datasets/CAiRE/ASCEND.
- Removed 嗯 呃 um uh
- Resolved [UNK]'s using whisper-medium
Dataset Details
Dataset Description
- Language(s): English, Simplified Chinese, Mixed
- License: Creative Common Attribution Share-Alike 4.0 International (CC-BY-SA 4.0)
Dataset Creation
Source Data
https://huggingface.co/datasets/CAiRE/ASCEND
Data Collection and Processing
- Load from source
from datasets import load_dataset, Audio as DSAudio
data_raw = load_dataset("CAiRE/ASCEND")
data_raw = data_raw.cast_column("audio", DSAudio(sampling_rate=16000))
- Clean stop words
import re
def clean_transcripts(x):
cjk = "[\u3400-\u4db5\u4e00-\u9fa5\u9fa6-\u9fbb\uf900-\ufa2d\ufa30-\ufa6a\ufa70-\ufad9\uff00-\uffef\u2e80-\u2eff\u3000-\u303f\u31c0-\u31ef\u2f00-\u2fdf\u2ff0-\u2fff\u3100-\u312f\u31a0-\u31bf\ufe10-\ufe1f\ufe30-\ufe4f\u2600-\u26ff\u2700-\u27bf\u3200-\u32ff\u3300-\u33ff]"
x = re.sub(r'\.\.\.|\s|^|$', ' ', x) # expanding space allows matching " uh uh" case
x = re.sub(rf"({cjk}|\s)([Uu][mh]|U[MH])({cjk}|\s)", r"\1 \3", x) # replace any uh surrounded by cjk or space
x = x.replace('嗯', ' ')
x = x.replace('呃', ' ')
x = re.sub(r"\s+", " ", x)
return x.strip()
data = data_raw.map(lambda x: {"transcription": clean_transcripts(x['transcription'])})
data = data.filter(lambda x: x["transcription"] != "")
- Isolate samples with UNKs
unks = data.filter(lambda x: "[UNK]" in x["transcription"])
unks.shape
{'train': (402, 9), 'test': (36, 9), 'validation': (63, 9)}
- Load whisper model. For Chinese, medium performs best.
from stable_whisper import load_faster_whisper
model = load_faster_whisper(
"medium",
device="cuda",
compute_type="float16",
)
- Resolve UNKs with whisper-medium
from sacrebleu.tokenizers.tokenizer_zh import TokenizerZh
from whisper_normalizer.basic import BasicTextNormalizer
import cn2an
import json
import jiwer
from tqdm.auto import tqdm
sacretok = TokenizerZh()
whisper_norm = BasicTextNormalizer()
def compute_mer(hyp, ref):
def norm(x):
return sacretok(cn2an.transform(whisper_norm(x), "an2cn"))
return jiwer.process_words(norm(hyp), norm(ref)).wer * 100
adjusted = {split:dict() for split in data}
double_check = {split:dict() for split in data}
UNK = "[UNK]"
for split in data:
trange = tqdm(unks[split], desc=split)
for i,sample in enumerate(trange):
transcription = sample['transcription']
texts = transcription.split(UNK)
words = []
for sent in texts[1:]:
for w in sacretok(sent).split():
if w not in words:
words += [w]
keyword = "关键词"
header = "字幕"
prompt = f"{keyword} \"{'/'.join(words)}\" {header} "
result = model.transcribe_stable(
audio=sample['audio']['array'],
initial_prompt=prompt, # encourage reuse of words
prefix=texts[0], # forcing start to follow real start
language=sample['language'].replace('mixed', 'zh'),
regroup=False,
verbose=None,
no_speech_threshold=1.0,
suppress_silence=False,
word_timestamps=True # though unused, timestamps reduce hallucination
).merge_all_segments()
adjustment = clean_transcripts(
result.text
.replace(keyword, " ")
.replace(header, " ")
)
mer=compute_mer(transcription, adjustment)
adjusted[split][sample['id']] = adjustment
trange.set_postfix(mer=f"{mer:.2f}", dc=len(double_check[split]))
if mer > 30:
double_check[split][sample['id']] = mer
print(transcription, "||", adjustment)
if i % 5 == 0 or i == len(unks[split]) - 1:
with open(f"checkpoint_{split}.json", "w") as f:
json.dump(adjusted[split], f)
- Replace UNK utterances with resolved ones
import json
adjusted_transcripts = {}
for split in data_raw:
with open(f"checkpoint_{split}.json", "r", encoding="utf8") as f:
adjusted_transcripts[split] = json.load(f)
UNK = "[UNK]"
def fix_unk(sample):
def bad(orig, new):
return sacretok(new) in sacretok(orig)
transcription = clean_transcripts(sample['transcription'].replace(UNK, ""))
sid = sample['id']
adjustment = adjusted_transcripts[split].get(sid, transcription)
if bad(transcription, adjustment):
# adjustment worse than just removing UNK
# print("skipped:", transcription, "||", adjustment)
adjustment = transcription
return adjustment
data = data_raw.map(lambda x: {"transcription": fix_unk(x)})
data = data.filter(lambda x: x["transcription"] != "")
data = data.sort(["session_id","id"])
for split in data:
for line in data[split]['transcription']:
assert UNK not in line
train adjusted 402 samples, 75 of which just removes UNKs.
test adjusted 36 samples, 9 of which just removes UNKs.
validation adjusted 63 samples, 7 of which just removes UNKs.