wav2vec2-large-xlsr-53-maltese-64h

The "wav2vec2-large-xlsr-53-maltese-64h" is an acoustic model suitable for Automatic Speech Recognition in Maltese. It is the result of fine-tuning the model "facebook/wav2vec2-large-xlsr-53" with around 64 hours of Maltese data developed by the MASRI Project at the University of Malta between 2019 and 2021. Most of the data is available at the the MASRI Project homepage https://www.um.edu.mt/projects/masri/.

The specific list of corpora used to fine-tune the model is:

• MASRI-HEADSET v2 (6h39m)
• MASRI-Farfield (9h37m)
• MASRI-Booths (2h27m)
• MASRI-MEP (1h17m)
• MASRI-COMVO (7h29m)
• MASRI-TUBE (13h17m)
• MASRI-MERLIN (25h18m) *Not available at the MASRI Project homepage

The fine-tuning process was perform during November (2022) in the servers of the Language and Voice Lab (https://lvl.ru.is/) at Reykjavík University (Iceland) by Carlos Daniel Hernández Mena.

Evaluation

import torch
from transformers import Wav2Vec2Processor
from transformers import Wav2Vec2ForCTC

#Load the processor and model.
MODEL_NAME="carlosdanielhernandezmena/wav2vec2-large-xlsr-53-maltese-64h"
processor = Wav2Vec2Processor.from_pretrained(MODEL_NAME)
model = Wav2Vec2ForCTC.from_pretrained(MODEL_NAME)

#Load the dataset
from datasets import load_dataset, load_metric, Audio
ds=load_dataset("common_voice", "mt", split="test")

#Normalize the transcriptions
import re
chars_to_ignore_regex = '[\\,\\?\\.\\!\\\;\\:\\"\\“\\%\\‘\\”\\�\\)\\(\\*)]'
def remove_special_characters(batch):
    batch["sentence"] = re.sub(chars_to_ignore_regex, '', batch["sentence"]).lower()
    return batch
ds = ds.map(remove_special_characters)

#Downsample to 16kHz
ds = ds.cast_column("audio", Audio(sampling_rate=16_000))

#Process the dataset
def prepare_dataset(batch):
    audio = batch["audio"]
    #Batched output is "un-batched" to ensure mapping is correct
    batch["input_values"] = processor(audio["array"], sampling_rate=audio["sampling_rate"]).input_values[0]
    with processor.as_target_processor():
        batch["labels"] = processor(batch["sentence"]).input_ids
    return batch
ds = ds.map(prepare_dataset, remove_columns=ds.column_names,num_proc=1)

#Define the evaluation metric
import numpy as np
wer_metric = load_metric("wer")
def compute_metrics(pred):
    pred_logits = pred.predictions
    pred_ids = np.argmax(pred_logits, axis=-1)
    pred.label_ids[pred.label_ids == -100] = processor.tokenizer.pad_token_id
    pred_str = processor.batch_decode(pred_ids)
    #We do not want to group tokens when computing the metrics
    label_str = processor.batch_decode(pred.label_ids, group_tokens=False)
    wer = wer_metric.compute(predictions=pred_str, references=label_str)
    return {"wer": wer}

#Do the evaluation (with batch_size=1)
model = model.to(torch.device("cuda"))
def map_to_result(batch):
    with torch.no_grad():
        input_values = torch.tensor(batch["input_values"], device="cuda").unsqueeze(0)
        logits = model(input_values).logits
    pred_ids = torch.argmax(logits, dim=-1)
    batch["pred_str"] = processor.batch_decode(pred_ids)[0]
    batch["sentence"] = processor.decode(batch["labels"], group_tokens=False)
    return batch
results = ds.map(map_to_result,remove_columns=ds.column_names)

#Compute the overall WER now.
print("Test WER: {:.3f}".format(wer_metric.compute(predictions=results["pred_str"], references=results["sentence"])))

Test Result: 0.011

BibTeX entry and citation info

When publishing results based on these models please refer to:

@misc{mena2022xlrs53maltese,
      title={Acoustic Model in Maltese: wav2vec2-large-xlsr-53-maltese-64h.}, 
      author={Hernandez Mena, Carlos Daniel},
      url={https://huggingface.co/carlosdanielhernandezmena/wav2vec2-large-xlsr-53-maltese-64h},
      year={2022}
}

Acknowledgements

The MASRI Project is funded by the University of Malta Research Fund Awards. We want to thank to Merlin Publishers (Malta) for provinding the audiobooks used to create the MASRI-MERLIN Corpus.

Special thanks to Jón Guðnason, head of the Language and Voice Lab for providing computational power to make this model possible. We also want to thank to the "Language Technology Programme for Icelandic 2019-2023" which is managed and coordinated by Almannarómur, and it is funded by the Icelandic Ministry of Education, Science and Culture.