Speech recognition with OpenSeq2Seq

Jason Li (NVIDIA), Vitaly Lavrukhin (NVIDIA)

2:30pm–3:10pm Thursday, October 31, 2019

Location: Grand Ballroom C/D

Text, language, speech

Average rating:

(5.00, 1 rating)

Who is this presentation for?

Researchers, engineers, and data scientists

Level

Intermediate

Description

Automatic speech recognition (ASR) is a core technology to create convenient human-computer interfaces. But building ASR systems with competitive word error rate (WER) traditionally required specialized expertise, large labeled datasets, and complex approaches.

Jason Li and Vitaly Lavrukhin dive into how end-to-end models simplified speech recognition and present Jasper, an end-to-end convolutional neural acoustic model, which yields state-of-the-art WER on LibriSpeech, an open dataset for speech recognition. They explore its implementation in the TensorFlow-based OpenSen2Seq toolkit and how to use it to solve large vocabulary speech recognition and speech command recognition problems. OpenSeq2Seq is an open source deep learning toolkit. They provide pretrained models for out-of-the-box experimentation.

Prerequisite knowledge

A basic understanding of deep learning and convolutional neural networks

What you'll learn

Discover end-to-end speech recognition and the OpenSeq2Seq deep learning toolkit

Jason Li

NVIDIA

Jason (Jing Yao) Li is a deep learning software engineer on the AI applications team at NVIDIA. He earned his BASc and MScAC at the University of Toronto working with Roger Grosse and Jimmy Ba. His research focus is on sequence-to-sequence models and speech, specifically in the domains of speech synthesis and speech recognition.

Vitaly Lavrukhin

NVIDIA

Vitaly Lavrukhin is a senior applied research scientist at NVIDIA, working on deep learning algorithms for speech and language technologies. Previously, he conducted research to solve computer vision problems with deep learning methods at Samsung R&D Institute Russia.