Adventures in voice (mis)recognition.

Let's get one thing straight. I am not actually bad at my job. Far from it, in fact. We have a set standard of accuracy that each respeaker much achieve before going online and transmitting captions. I have gone above and beyond that, with my last accuracy check registering at a very, very high level. But the nature of the beast is that variables like the weather and the subsequent condition of my sinuses, personal mood and level of alertness, the quality and nature of the material, the myriad kinks and quirks of the voice recognition programme we use, as well as the ongoing condition of the equipment we are given all combine to form a fluctuating 3D landscape full of various paths that can lead to anything from minor inaccuracies to monumental fuck-ups. 

I repeat what your newsreaders and sports commentators are saying as they are saying it. I caption live TV. Sometimes it's irritating and sometimes it's traumatic, but very rarely is it boring.

The thing with respeaking is that you always have to sound the same. My accuracy readings fluctuated wildly until I arrived at my proper captioning voice, which I call The Sexy Dalek. Monotonous but soft with rounded vowels. But beware - speak too robotically and you will get every word that ends in a plosive consonant being followed by by and extra "or"; breathe too deeply at the beginning of a sentence and it will begin with "if," "of," "five," or "four". Your sentences may read like something out of the wrong century: "Put on your wellies, for the weather today will be wet..." Or even worse, you could entice your viewer into a state of perpetual suspense: "If there were four victims in the accident..." "If police had released a picture of the gunman..." Thankfully, with a combination of training my own voice and employing some sneaky house styles and dictation macros, that sort of thing is being gradually reduced.

Your diction must be clear. Failure to enunciate each syllable can lead to slurring and some embarrassing errors. This is where voice recognition becomes voice misrecognition.

The most recent incident occurred a few days ago, when the words "artificial flowers" transmitted as "artifical phallus." This was awkward enough, but even more so when the "phallus" in question was being left as a reflection of grief at a loved one's passing... in a cemetery. Oops.


Stay tuned for more adventures in (mis)recognition, where your weather report can really turn apocalyptic ("Met Office warnings are in place for eyes on the road. Snow in some places, brain in others.").