Automated Transcription? Be Careful What You Wish For!

speech enabled email
Voice recognition technology has finally reached a level of sophistication matching the precision and native awareness of human transcription! Just check out the level of unbelievable naturalness offered by Google Voice, the search giant’s flagship automated voicemail transcription service-


“Hi Oleg, My name is patio and I got your message for focal stuff I’m selling,
please give me a call back… my number is XXXXXXXXX18 Thank God.”


“Thank God” ? Errr… ok, so maybe that’s not the best automated transcription possible, but it’s probably just a fluke, right? I mean… Google isn’t exactly known for releasing half-finished products onto the marketplace, are they? Let’s look at another automated transcription we recently received through Google Voice-


“Hello Oleg, This is Jennifer parking you…”


Woops! Probably just another fluke. Jennifer probably didn’t really say she was “parking” us, but maybe she was a soft talker or had an impediment or a strong accent, or maybe she left her message while walking through windy traffic. You can’t really expect an automated transcription service to deal with those sorts of obstacles, right?

Here, another voicemail transcription that just landed in my inbox, this one should be pretty accurate-


“Hello, this is. Crystaline filled Burger transfer and storage
Allied Van Lines calling for Oleg the beta about moving from Saint Paul,
Minnesota. 2. Germantown Maryland moving with taish Technologies I have
been asked by Tah to get an appointment set an on site surveys a walk through
that…You can reach me, please do call so we can get that appointments
it because it’s my understanding of us to move in the next couple of week.”


Can someone, please, hold my chair…

On Google’s Voicemail Transcriptions website they explain they are


“the only fully automated voicemail transcription on the market.”

According to Google, their frontrunner status means the service is “not perfect yet.”

I don’t know what Google is talking about! After all, any automated transcription service that’s able to accurately recognize about a concept as specialized as “Crystaline filled Burger transfer and storage” seems pretty perfect to us!

Heck, Google Voice was even able to figure out the caller has “been asked by Tah to get an appointment set an on site surveys a walk through that…” Sure, this might not be perfect English translation, but it didn’t take more than 10-15 minutes to figure out what that particular sentence might have potentially meant.

And saving time is what Google Voice’s automated transcription service is all about! Ok, ok, it might take 50 times as long to parse through one of these transcriptions as to actually just check our voicemail, but it could be much worse! A year ago it would have taken 100 times as long to translate an automated transcription than it takes today!

Besides, as Google points out, they don’t guarantee accuracy because their software doesn’t always produce absolutely perfect transcriptions. If everyone would just enunciate clearly, eliminate all background noise, and use high definition microphones when leaving voicemail, none of these problems would ever pop up! So don’t blame Google Voice for the above inaccuracies- it’s really the caller’s fault. Heck, if I could speak perfectly every time I leave voice mail and make the software close to 95% accurate they would figure a way to charge me for this service! After all, it would be almost perfect!

I mean, who cares if Google Voice can’t even accurately transcribe a plain-speak phone call from a real estate broker, this automated transcription service is absolutely ready to be used within the jargon-filled fields of legal and medical transcription, especially considering the life-or-death importance of precise language required by professionals within these fields.

After all, what working professional wouldn’t want to receive a transcription like the following in their inbox when communicating with an important client?

“Yo. This is Mary… I’m calling to say thank you for choosing Health Partners
and welcome. Whether you’re new in a while. Or it’s been a member for years. We’re
glad to be your partner in hell…”

Sure, that’s probably not exactly what Mary (or whatever her real name is) had to say, but at least Google Voice totally understands that we all need a good laugh every once in a while to help break up the workday.

Laughs aside, the problems with Google Voice is due to the fact that even clearly enunciated speech is incredibly complex and the complexity of transcribing speech increases exponentially when you introduce background noise or multiple speakers on the same call. In fact the only truly accurate form of transcription on the market remains human transcription but Google should probably be commended for taking on such a massive task and coming up with an ingenious way of enlisting our help.