wav2letter++ – end-to-end speech recognition system

Last Updated on July 13, 2021

wav2letter++ is a fast, open source speech processing toolkit from the Speech team at Facebook AI Research built to facilitate research in end-to-end models for speech recognition. It’s intended as a platform for rapid research in end-to-end speech recognition.

It follows a completely convolutional approach, and uses Convolutional Neural Networks (CNN) for acoustic modelling as well as language modelling.

wav2letter++ uses the ArrayFire tensor library and the flashlight machine learning library for maximum efficiency. ArrayFire is a highly optimized tensor library that can execute on multiple backends including a CUDA GPU back-end and a CPU backend. ArrayFire also uses just-in-time code generation to combine series of simple operations into a single kernel call. This results in faster execution for memory bandwidth bound operations and can reduce peak memory use.

wav2letter has been moved and consolidated into Flashlight in the ASR application.

wav2letter++ is free and open source software.

Features include:

• Supports three main modes of training: train (flat-start training), continue (continuing with a checkpoint state), and fork (for e.g. transfer learning). The training pipeline scales seamlessly using data-parallel, synchronous stochastic-gradient-descent as well as inter process communications powered by NVIDIA Collective Communication Library. Training times scale linearly to 64 GPUs with 100 million parameters.
• Supports several end-to-end sequence models as well as a wide range of network architectures and activation functions.
• Supports feature extraction across different audio formats. The framework computes features on the fly prior to each network evaluation and enforces asynchrony and parallelization in order to maximize efficiency during the training of models.
• Training pipeline gives maximum flexibility for the user to experiment with different features, architectures and optimization parameters.
• Decoder is a beam-search decoder with several optimizations to improve efficiency.
• Written entirely in C++ which makes it efficient to train models and perform real-time decoding.

Website: github.com/facebookresearch/wav2letter/wiki
Support: Tutorials, GitHub Code Repository
Developer: Facebook, Inc. and its affiliates
License: BSD License

wav2letter++ is written in C++ and Python. Learn C++ with our recommended free books and free tutorials. Learn Python with our recommended free books and free tutorials.

Return to Speech Recognition Tools Home Page

Explore Linux
	Best Free and Open Source Software - A huge collection of recommended free and open source software covering every major category.
	In-depth Linux Software Reviews - Detailed assessments of Linux software covering features, usability and overall quality.
	The Big List of Active Linux Distros - An extensive collection of actively developed Linux distributions.
	Open Source Alternatives - Alternatives to software and services from Google, Microsoft, Apple, Adobe, IBM, Autodesk, Oracle and others.
	Essential Linux System Tools - Indispensable utilities for monitoring, maintaining and managing Linux systems.
	Linux Productivity Tools - Useful utilities for improving productivity and streamlining everyday tasks.
	Alternatives to Popular CLI Tools - Modern alternatives to familiar command-line utilities supplied with Linux systems.
	Awesome Free Linux Game Tools - Tools, utilities and companion software that improve gaming on Linux.
	Machine Learning on Linux - Machine learning and deep learning software, self-hosted applications and practical tools.
	Saving Money with Linux - Practical ways to reduce electricity consumption and lower computing costs with Linux.
	Linux Candy - Fun, entertaining and sometimes delightfully quirky software from the lighter side of Linux.