audio-kw-in-context.py

# Copyright 2020 The HuggingFace Datasets Authors and the current dataset script contributor.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""sil-ai/audio-kw-in-context is a subset of MLCommons/ml_spoken_words focusing on keywords found in the Bible'"""

import json
import os

import datasets

_CITATION = """\
@InProceedings{huggingface:audio-kw-in-context,
title = {audio-kw-in-context},
author={Joshua Nemecek
},
year={2022}
}
"""
_DESCRIPTION = 'sil-ai/audio-kw-in-context is a subset of MLCommons/ml_spoken_words focusing on keywords found in the Bible'
_LANGUAGES = ['eng', 'ind', 'spa']
_LANG_ISO_DICT = {'en':'eng','es':'spa','id':'ind'}
_HOMEPAGE = 'https://ai.sil.org'
_URLS = {"metadata": "bible-keyword-context.json",
         "files": {lang: f'https://audio-keyword-spotting.s3.amazonaws.com/HF-context-v2/{lang}-kw-archive.tar.gz' for lang in _LANGUAGES},
         }
_LICENSE = 'CC-BY 4.0'
_GENDERS = ["MALE", "FEMALE", "OTHER", "NAN"]

class AudioKwInContextConfig(datasets.BuilderConfig):
    """BuilderConfig for Audio-Kw-In-Context"""
    def __init__(self, language='', **kwargs):
        super(AudioKwInContextConfig, self).__init__(**kwargs)
        self.language = _LANG_ISO_DICT.get(language, language)

class AudioKwInContext(datasets.GeneratorBasedBuilder):
    """Audio-Kw-In-Context class"""
    BUILDER_CONFIGS = [AudioKwInContextConfig(name=x, description=f'Audio keyword spotting for language code {x}', language=x) for x in _LANGUAGES]
    
    DEFAULT_CONFIG_NAME = ''

    BUILDER_CONFIG_CLASS = AudioKwInContextConfig

    VERSION = datasets.Version("0.0.2")

    def _info(self):
        features = datasets.Features(
                {
                    "file": datasets.Value("string"),
                    "language": datasets.Value("string"),
                    "speaker_id": datasets.Value("string"),
                    "sentence": datasets.Value("string"),
                    "keywords": datasets.Sequence(datasets.Value("string")),  
                    "audio": datasets.Audio(sampling_rate=16_000),
                    "start_times": datasets.Sequence(datasets.Value("float32")),
                    "end_times": datasets.Sequence(datasets.Value("float32")),
                    "confidence": datasets.Sequence(datasets.Value("float32")),
                }
        )

        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=features,  
            homepage=_HOMEPAGE,
            license=_LICENSE,
            citation=_CITATION,
        )

    def _split_generators(self, dl_manager):
        
        if self.config.language == '':
            raise ValueError('Please specify a language.')
        elif self.config.language not in _LANGUAGES:
            raise ValueError(f'{self.config.language} does not appear in the list of languages: {_LANGUAGES}')

        data_dir = dl_manager.download(_URLS['metadata'])
        with open(data_dir, 'r') as f:
            filemeta = json.load(f)

        audio_dir = dl_manager.download_and_extract(_URLS['files'][self.config.name])

        langmeta = filemeta[self.config.language]

        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN,
                # These kwargs will be passed to _generate_examples
                gen_kwargs={
                    "audio_dir": audio_dir,
                    "data": langmeta,
                    "split": "train",
                },
            ),
            datasets.SplitGenerator(
                name=datasets.Split.VALIDATION,
                # These kwargs will be passed to _generate_examples
                gen_kwargs={
                    "audio_dir": audio_dir,
                    "data": langmeta,
                    "split": "dev",
                },
            ),
            datasets.SplitGenerator(
                name=datasets.Split.TEST,
                # These kwargs will be passed to _generate_examples
                gen_kwargs={
                    "audio_dir": audio_dir,
                    "data": langmeta,
                    "split": "test",
                },
            ),
        ]

    # method parameters are unpacked from `gen_kwargs` as given in `_split_generators`
    def _generate_examples(self, audio_dir, data, split):
        for key, row in enumerate(data[split]):
            try:
                trows = row['file'].split('/')
                tfile = os.path.join(audio_dir, trows[-1])
                if not tfile.endswith('.mp3'):
                    os.rename(tfile, tfile + '.mp3')
                    tfile += '.mp3'
                yield key, {
                    "file": tfile,
                    "sentence": row.get('sentence'),
                    "language": self.config.language,
                    "speaker_id": row.get('speaker_id',row.get('client_id')),
                    "keywords": row['keywords'],  
                    "audio": tfile,
                    "start_times": row.get('start_times'),
                    "end_times": row.get('end_times'),
                    "confidence": row.get('confidence'),
                }
            except Exception as e:
                print(e)
                print(f'In split {split}: {row["file"]} failed to load. Data may be missing.')
                pass