language:
- en library_name: nemo datasets:
- librispeech_asr
- fisher_corpus
- Switchboard-1
- WSJ-0
- WSJ-1
- National-Singapore-Corpus-Part-1
- National-Singapore-Corpus-Part-6
- vctk
- VoxPopuli-(EN)
- Europarl-ASR-(EN)
- Multilingual-LibriSpeech-(2000-hours)
- mozilla-foundation/common_voice_7_0 thumbnail: null tags:
- automatic-speech-recognition
- speech
- audio
- CTC
- Conformer
- Transformer
- pytorch
- NeMo
- hf-asr-leaderboard
- Riva license: cc-by-4.0
This speech tagger performs transcription for Hindi, annotates key entities, predict speaker age, dialiect and intent.
Model is suitable for voiceAI applications, real-time and offline.
Model Details
- Model type: NeMo ASR
- Architecture: Conformer CTC
- Language: English
- Training data: CommonVoice, Gigaspeech
- Performance metrics: [Metrics]
Usage
To use this model, you need to install the NeMo library:
pip install nemo_toolkit
How to run
import nemo.collections.asr as nemo_asr
# Step 1: Load the ASR model from Hugging Face
model_name = 'WhissleAI/stt_hi_conformer_ctc_entities_age_dialiect_intent'
asr_model = nemo_asr.models.EncDecCTCModel.from_pretrained(model_name)
# Step 2: Provide the path to your audio file
audio_file_path = '/path/to/your/audio_file.wav'
# Step 3: Transcribe the audio
transcription = asr_model.transcribe(paths2audio_files=[audio_file_path])
print(f'Transcription: {transcription[0]}')
Dataset is from AI4Bharat IndicVoices Hindi V1 and V2 dataset.