Accent unplugged: Evaluating Voice assistants

Have you ever been curious about why Voice Assistants (Siri, Amazon Echo and Google Assistant) sometimes struggle to understand your English accent? As linguists, we can play a crucial role in evaluating these sophisticated systems. 

In this class, you will assess the performance of Automatic Speech Recognition systems (ASR) by evaluating them against your own recorded speech and other existing speech recordings. You will learn how to use recording equipment in a laboratory setting as well as how to cut and prepare audio for further processing.

Throughout the course, you will gain the knowledge in the following areas: 

• Fundamentals of ASR
• Techniques of making recordings in a laboratory setting
• Dealing with audios
• Evaluating ASRs with your own voice
• Assessing ASRs on other English accent speech recordings

The AP is a documentation of the work you have done in the class. As you progress through the course, there will be writing tasks that summarize the work you have done in the previous session(s). Using the assignment feedback, you can draft the relevant sections of your term paper. The final AP will then comprise an improved and expanded version of these write-ups into a 6-7 page term paper.

Pre-requisites: None, except the willingness and enthusiasm to learn about language technologies and linguistics. Please bring your laptop for each session.