Open-source Dutch Speech Recognition Model xlsr300m_cv_8.0_nl - Free Automatic Speech-to-Text Conversion

Xlsr300m Cv 8.0 Nl

Developed by Iskaj

This is a Dutch automatic speech recognition (ASR) model based on the XLS-R-300M architecture, trained on the Common Voice 8.0 Dutch dataset.

Speech Recognition

Transformers

OtherOpen Source License:Apache-2.0 #Dutch Speech Recognition #Multi-scenario Robustness #Low CER Performance

Downloads 33

Release Time : 3/2/2022

Model Overview

This model is specifically designed for Dutch automatic speech recognition tasks, capable of converting Dutch speech into text.

Model Features

Multi-dataset Training

Trained on Common Voice 8.0 and 7.0 Dutch datasets, enhancing the model's generalization capability.

Robust Speech Recognition

Capable of processing Dutch speech with various accents and background noise.

Efficient Inference

The 300M parameter size provides relatively efficient inference speed while maintaining good performance.

Model Capabilities

Dutch Speech-to-Text

Continuous Speech Recognition

Multi-accent Recognition

Use Cases

Speech Transcription

Dutch Meeting Minutes

Automatically transcribe Dutch meeting recordings into text records.

WER of 46.94% on the Common Voice test set.

Voice Assistant

Used as the speech recognition module for Dutch voice assistants.

WER of 42.56% on the robust speech event test set.

Education

Language Learning Applications

Help learners practice Dutch pronunciation and listening.

🚀 xlsr300m_cv_8.0_nl

This is a model for automatic speech recognition in Dutch, offering evaluation commands and inference examples.

🚀 Quick Start

Evaluation Commands

To evaluate on mozilla-foundation/common_voice_8_0 with split test

python eval.py --model_id Iskaj/xlsr300m_cv_8.0_nl --dataset mozilla-foundation/common_voice_8_0 --config nl --split test

To evaluate on speech-recognition-community-v2/dev_data

python eval.py --model_id Iskaj/xlsr300m_cv_8.0_nl --dataset speech-recognition-community-v2/dev_data --config nl --split validation --chunk_length_s 5.0 --stride_length_s 1.0

💻 Usage Examples

Basic Usage

import torch
from datasets import load_dataset
from transformers import AutoModelForCTC, AutoProcessor
import torchaudio.functional as F

model_id = "Iskaj/xlsr300m_cv_8.0_nl"

sample_iter = iter(load_dataset("mozilla-foundation/common_voice_8_0", "nl", split="test", streaming=True, use_auth_token=True))

sample = next(sample_iter)
resampled_audio = F.resample(torch.tensor(sample["audio"]["array"]), 48_000, 16_000).numpy()

model = AutoModelForCTC.from_pretrained(model_id)
processor = AutoProcessor.from_pretrained(model_id)

inputs = processor(resampled_audio, sampling_rate=16_000, return_tensors="pt")
with torch.no_grad():
  logits = model(**inputs).logits
  predicted_ids = torch.argmax(logits, dim=-1)
  transcription = processor.batch_decode(predicted_ids)

transcription[0].lower()
#'het kontine schip lag aangemeert in de aven'

📚 Documentation

Model Information

Property	Details
Language	Dutch
Tags	automatic-speech-recognition, mozilla-foundation/common_voice_8_0, mozilla-foundation/common_voice_7_0, nl, robust-speech-event, model_for_talk, hf-asr-leaderboard
Datasets	mozilla-foundation/common_voice_8_0

Model Index

Name: XLS - R - 300M - Dutch
Results:
- Task: Automatic Speech Recognition
  - Dataset: Common Voice 8 NL (mozilla-foundation/common_voice_8_0, nl)
    - Metrics:
      - Test WER: 46.94
      - Test CER: 21.65
- Task: Automatic Speech Recognition
  - Dataset: Robust Speech Event - Dev Data (speech-recognition-community-v2/dev_data, nl)
    - Metrics:
      - Test WER: ???
      - Test CER: ???
- Task: Automatic Speech Recognition
  - Dataset: Robust Speech Event - Test Data (speech-recognition-community-v2/eval_data, nl)
    - Metrics:
      - Test WER: 42.56

📄 License

This project is licensed under the Apache - 2.0 license.

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご