Vonage Visual Voicemail is a funny thing. And I don’t mean funny in the sense of “weird,” either. I mean funny in the sense of “I’m cracking up right now.” (Go right to the examples.)
Why do I say that? Because it sometimes outputs utterly ridiculous transcriptions. Can I fault Vonage for its less-than-perfect speech-to-text processing? Yes and no.
Vonage, the Voice over Internet Protocol (VoIP) provider, offers a free transcription service to its customers. Subscribers can sign up for free email delivery to receive text versions of callers’ voicemail messages.
I like Vonage; in fact, I’ve been using it for over ten years. Why? Mainly because it offers very affordable rates on long-distance calling when I want the relative reliability of a land line. (It’s not really a land line, of course. But the call never drops unless you’re getting spotty service from your cable Internet provider).
However, I do not use Vonage for the accuracy of its voicemail transcriptions.
The Limitations of Vonage Visual Voicemail
For all the mellifluous alliteration of its name, Vonage Visual Voicemail is flawed. Take the following, for example: a native speaker of French called RedLine and left a message in English. The speaker has a fairly strong accent, but Vonage tried its best to produce a usable transcription anyway.
There are two columns below. On the left in red is the transcription that I received from Vonage Visual Voicemail. In blue is what the speaker actually said. There was more wrong than right in Vonage’s output. In addition, some of it was laugh-out-loud ridiculous.
Vonage Visual Voicemail
And sign so much for watching the today’s lesson, five, six, I mean yeah Bridget it’s mom that’s pretty new but it’s all good. I love you. I’ve been to the old out.
And thanks so much for what you did today. It was fantastic. I […] many apologies for not calling you back before, but I was “live” until now.
Wait a sec—did my client just tell me that she loves me? No. She’s a journalist who was covering a speaker, so she was “live” (attending a live press conference).
I’m that’s one at the farm. So I’ll see you in the afternoon. So he starts line once you walk at far south in the afternoon. That’s pretty big red.
[…] and one at four-thirty in the afternoon. So if your friend wants to work at four-thirty in the afternoon, that will be great.
“That’s pretty big red” is my new favorite phrase. Utterly meaningless, but still my new favorite.
That are pretty to me to work for you tomorrow morning. I just saw how it’s Ross Junior it’s Deb. I need to boost your model one in the morning.
[…] and hopefully it will work for you tomorrow morning and for her tomorrow afternoon. There are only two pools tomorrow, one in the morning […]
I have no idea how Ross Junior and Deb got into my voicemail message, but I want them out.
My Olive Branch to Vonage
In fairness to Vonage, the speaker’s accent may have even posed a problem to a listener unfamiliar with a thick French accent. And the company’s transcriptions are usually better than the example here.
Take a look at the message below. The Vonage Visual Voicemail notification that I received looked like this (I’ve redacted confidential information):
Hi this message is for Matthew my name is [redacted] I’m calling you from [redacted]. We have a requirement for [redacted]. And we wanted to reach out to you to see your availability to maybe assist us with some of those training some of them start as early as April so if you could please give me a call back so we can discuss and see if you can help us with some of. My number is [redacted]. Thank you have a good day.
So Vonage can certainly produce a very usable transcription. I just don’t know how often that happens. When it can’t transcribe your voicemail message, Vonage tells you the following: We’re sorry. We were unable to transcribe this message. Please check your voicemail.
Botch jobs like “I’m that’s one at the farm” highlight just how far Vonage’s technology has to go in order to become a consistently reliable tool for customers.
Speech-to-text technologies have to perform an inherently difficult task. But a user may reasonably question the benefit of such a service when the result is so poor.
If you have a funny transcription to share, then tell me about it in the comments!