|
|
|
|
|
|
|
|
|
|
|
import json |
|
|
|
import datasets |
|
|
|
_DESCRIPTION = """\ |
|
Grapheme-to-Phoneme training, validation and test sets |
|
""" |
|
|
|
_BASE_URL = "https://huggingface.co/datasets/flexthink/librig2p-nostress-space/resolve/main/dataset" |
|
|
|
_HOMEPAGE_URL = "https://huggingface.co/datasets/flexthink/librig2p-nostress-space" |
|
|
|
_ORIGINS = ["librispeech", "librispeech-lex", "wikipedia-homograph"] |
|
_NA = "N/A" |
|
_SPLIT_TYPES = ["train", "valid", "test"] |
|
_DATA_TYPES = ["lexicon", "sentence", "homograph"] |
|
_SPLITS = [ |
|
f"{data_type}_{split_type}" |
|
for data_type in _DATA_TYPES |
|
for split_type in _SPLIT_TYPES] |
|
|
|
class GraphemeToPhoneme(datasets.GeneratorBasedBuilder): |
|
def __init__(self, base_url=None, splits=None, *args, **kwargs): |
|
super().__init__(*args, **kwargs) |
|
self.base_url = base_url or _BASE_URL |
|
self.splits = splits or _SPLITS |
|
|
|
def _info(self): |
|
return datasets.DatasetInfo( |
|
description=_DESCRIPTION, |
|
features=datasets.Features( |
|
{ |
|
"id": datasets.Value("string"), |
|
"speaker_id": datasets.Value("string"), |
|
"origin": datasets.Value("string"), |
|
"char": datasets.Value("string"), |
|
"phn": datasets.Sequence(datasets.Value("string")), |
|
"homograph": datasets.Value("string"), |
|
"homograph_wordid": datasets.Value("string"), |
|
"homograph_char_start": datasets.Value("int32"), |
|
"homograph_char_end": datasets.Value("int32"), |
|
"homograph_phn_start": datasets.Value("int32"), |
|
"homograph_phn_end": datasets.Value("int32"), |
|
}, |
|
), |
|
supervised_keys=None, |
|
homepage=_HOMEPAGE_URL, |
|
) |
|
|
|
def _get_url(self, split): |
|
return f'{self.base_url}/{split}.json' |
|
|
|
def _split_generator(self, dl_manager, split): |
|
url = self._get_url(split) |
|
path = dl_manager.download_and_extract(url) |
|
return datasets.SplitGenerator( |
|
name=split, |
|
gen_kwargs={"datapath": path, "datatype": split}, |
|
) |
|
|
|
def _split_generators(self, dl_manager): |
|
return [ |
|
self._split_generator(dl_manager, split) |
|
for split in self.splits |
|
] |
|
|
|
def _generate_examples(self, datapath, datatype): |
|
with open(datapath, encoding="utf-8") as f: |
|
data = json.load(f) |
|
breakpoint() |
|
for sentence_counter, (item_id, item) in enumerate(data.items()): |
|
resp = { |
|
"id": item_id, |
|
"speaker_id": str(item.get("speaker_id") or _NA), |
|
"origin": item["origin"], |
|
"char": item["char"], |
|
"phn": item["phn"], |
|
"homograph": item.get("homograph", _NA), |
|
"homograph_wordid": item.get("homograph_wordid", _NA), |
|
"homograph_char_start": item.get("homograph_char_start", 0), |
|
"homograph_char_end": item.get("homograph_char_end", 0), |
|
"homograph_phn_start": item.get("homograph_phn_start", 0), |
|
"homograph_phn_end": item.get("homograph_phn_end", 0) |
|
|
|
} |
|
yield sentence_counter, resp |
|
|
