Training your first ASR model: Introduction to ASR in linguistic research
Abstract
This workshop introduces the inner workings of Automatic Speech Recognition (ASR) and the classical ASR architecture. It focuses on ASR practices that facilitate linguistic research and provides a flexible workflow of automatic forced alignment, demonstrated through various research scenarios. The workshop aims to help you understand the basic concepts in ASR and guide you to utilise ASR in your own linguistic research.
Prerequisite:
The workshop will be easier for you if you have basic knowledge of speech processing, Unix Shell, and Python. But you can definitely still benefit if you do not have such knowledge.
Pre-workshop Setup:
We will mainly be using Python and Montreal Forced Aligner (MFA), which can be installed through Anaconda or Miniconda.
For Windows, here’s a tutorial by Dr Cong Zhang for the installation and usage of MFA 2.0.
For Mac, please follow the installation guide on MFA website.
Dr Chenzi Xu
Leverhulme Early Career Fellow
My research interests include speech prosody, speech perception, and speech technology.