Speak the command shown below as naturally as you would to a real voice assistant. After each command, you'll be shown what the system understood, and asked whether it's correct.
Before you begin
Your spoken commands are transcribed by one of three ASR models per trial, chosen at random — you won't be told which.
Your recorded audio is used only to generate a transcript and is deleted immediately afterward. It is never stored, saved, or shared in any form.
We collect your gender, age group, and (optionally) first language to check whether system performance is consistent across groups. No other personal information is collected.
Responses are logged under a random session ID, not your name.
Since your audio is deleted right after transcription, this is the only record of background noise conditions for the analysis.
COMMAND 1 OF 9
Read this aloud
—
Click to record
Routing to blind condition & analyzing…
Command type (intent)—
Classifier of command (NER)—
Was the command type correct?
Was the detected item correct?
Quick feedback
Before you go, please rate how much you agree with each statement about the system you just used (1 = Strongly disagree, 7 = Strongly agree).