Show HN: The worst voicemail transcription ever

8 points by SQL2219 ↗ HN
How did this make it to Production?

Outlook - Voice Mail Preview:

Hey dad this is Darcy with your software and we can I was just checking on you I'm looking into how things are going and also sort of a quick housekeeping question it looks like you might be a version or show behind on the fries manager in comparison to shut let me know if you need any information you know stop being shipment so I changed troll with me that.

3 comments

[ 4.2 ms ] story [ 16.2 ms ] thread
I've got another contender.

From my Grandpa, transcribed by Google Voice:

Hey,<redacted first name>, It's Pam on Pat be just driving up to the lake with his check in and see if you had a nice thanksgiving and enjoy the parade yesterday, and you were missed you had a nice thanksgiving with Kevin ward and it's only for a couple days just grandma died so talk to you later, okay. Alright

Relevant part: "it's only for a couple days just grandma died so talk to you later, okay."

Voice transcription is hard.

Think of how many accents there exist "in the wild". About the same as the number of people in the world.

That is true, made more so when doing phone calls (8kHz, Mono recordings) with a variety of accents. I've scored a live transcription system at 94% accurate over 250 samples of voicemail recordings from UK accents, with various automated services coming in at the following rates:

AWS 0.871603 Temi 0.871063 Google 0.867890 Speechmatics 0.858851 OtterAI 0.855769 Microsoft 0.854067 AssemblyAI 0.795696 IBM 0.777215 Remeeting 0.771764

I'm building a new transcription startup called 3Scribe (shameless self-promotion) that's scoring at just below 91%. Hoping to squeeze out another little bit of accuracy but at the moment struggling with React-Bootstrap, trying to build a beta MVP.