This collection contains medium size versions of Conformer-CTC (around 30M parameters) trained on ULCA Marathi Corpus with around 1300 hours of marathi speech. The model transcribes speech in marathi characters along with spaces.
Conformer-CTC model is a non-autoregressive variant of Conformer model  for Automatic Speech Recognition which uses CTC loss/decoding instead of Transducer. You may find more info on the detail of this model here: Conformer-CTC Model
The tokenizers for these models were built using the text transcripts of the train set with this script.
The checkpoint of the language model used as the neural rescorer can be found here. You may find more info on how to train and use language models for ASR models here: ASR Language Modeling
All the models in this collection are trained on ULCA Marathi Labelled Dataset (~1300 hrs)
The tokenizer for this model was built using text corpus provided with the train dataset.
We build a token set with the following script:
python [NEMO_GIT_FOLDER]/scripts/tokenizers/process_asr_text_tokenizer.py \ --manifest="train_manifest.json" \ --data_root="" \ --vocab_size=512 \ --tokenizer="spe" \ --spe_type="unigram" \ --spe_character_coverage=1.0 \ --log
Performance The list of the available models in this collection is shown in the following table. Performances of the ASR models are reported in terms of Word Error Rate (WER%) with greedy decoding and 6-Gram KenLM trained on AI4Bharat Corpus.
6-Gram KenLM with 128 beam size with n_gram_alpha=1.5, n_gram_beta=2.0.
Greedy Decoding Scores:
The model is available for use in the NeMo toolkit , and can be used as a pre-trained checkpoint for inference or for fine-tuning on another dataset.
import nemo.collections.asr as nemo_asr asr_model = nemo_asr.models.EncDecCTCModelBPE.from_pretrained(model_name="stt_mr_conformer_ctc_medium")
python [NEMO_GIT_FOLDER]/examples/asr/transcribe_speech.py \ pretrained_name="stt_mr_conformer_ctc_medium" \ audio_dir=""
This model accepts 16000 KHz Mono-channel Audio (wav files) as input.
This model provides transcribed speech as a string for a given audio sample.
Since this model was trained on publically available speech datasets, the performance of this model might degrade for speech which includes technical terms, or vernacular that the model has not been trained on. The model might also perform worse for accented speech.