Abstract

Speech recognition has been around for 70 years, but has come a long way in the last decade. Today, most people interact with speech recognition in some form every day. The current speech recognition approaches have evolved over the decades and incorporated knowledge from many different disciplines like computer science, signal processing, linguistics, statistics, and machine learning. This presentation will give an overview of how speech recognition has progressed over the years, and describe some of the key approaches/algorithms that have been used. Training and/or running your own speech recognition models has never been easier, and this talk will give some background on how we got here and why speech recognition is more accessible to researchers than it has ever been in the past.

About the Speaker

Stan Salvador is the Chief scientist at Cobalt Speech and Language where he works on various projects related to Speech Recognition and Natural Language Understanding. Previous work includes creating Echo/Alexa at Amazon, implementing voicemail transcription at a small startup company, developing algorithms to create schedules for petroleum pipelines, creating machine learning models for chemical weapon detectors, and working on a research project to detect valve failures in space shuttle engines at Florida Tech.