CommonVoice support for convert_hf_dataset_to_nemo script

[SeanNaren]

What does this PR do ?

Related #5638 #5772

Renames sentence column to text before creating manifest. Exposes ensure_ascii for saving the transcripts + fix naming of config class.

Collection: ASR

Changelog

• CommonVoice support for convert_hf_dataset_to_nemo script.

Usage

• You can potentially add a usage example below

python scripts/speech_recognition/convert_hf_dataset_to_nemo.py \
        output_dir=/data/commonvoice/ \
        path="mozilla-foundation/common_voice_11_0" \
        name="eo" \ # esperanto
        split="validation" \
        ensure_ascii=False \
        use_auth_token=True

Pre checks:

• Make sure you read and followed Contributor guidelines
• Did you write any new necessary tests?
• Did you add or update any necessary documentation?
• Does the PR affect components that are optional to install? (Ex: Numba, Pynini, Apex etc)
  • Reviewer: Does the PR have correct import guards for all optional libraries?

PR Type:

• New Feature
• Bugfix
• Documentation

[SeanNaren]

I ran this with ensure_ascii=True and ensure_ascii=False. I then ran the manifest with the below code:

from nemo.collections.asr.data.audio_to_text import ASRManifestProcessor
from nemo.collections.common.parts.preprocessing.parsers import CharParser


x = ASRManifestProcessor(
    manifest_filepath='/data/commonvoice/mozilla-foundation/common_voice_11_0/eo/validation/validation_mozilla-foundation_common_voice_11_0_manifest.json',
    parser=CharParser(['a','b','c'])
)

Both loaded successfully, seems no issue with the manifests @titu1994

[titu1994]

LGTM, thanks !