Sunday, 12 June 2016

Word Up: Speech-to-Text

In light of recently maiming myself, I decided to try out Word's speech-to-text function. 

Whilst partially incapacitated, I use my good hand to brows websites. On a news channel I stumbled across something that completely grabbed my attention. So much so that eight hours later I was still reading about it, going from Wiki to Wiki, jotting down notes. I think I've found my next major story subject, but it's driving me up the wall because I can't actually start writing about it. My right hand is currently doing the work of both hands, and it's exhausting. There is a saying: 

On balance, as annoying as it is not being able to tell my story, falling into a fire was a thousand times more agonising.

Anyway, Word.

I've been a huge fan of text-to-speech for a long time. I regularly get Word to read e-mails and news articles to me whilst I contemplate my belly fluff.

Unfortunately, speech-to-text ain't quite all there yet.

Here's a brief Windows explanation on turning on speech recognition for Windows 10 and some further information for Windows 8. Basically, in 10: File Explorer/This PC/Open Settings/type Speech into the Search Box/Speech Recognition.

It's worth taking the time to Train Your Computer, which basically involves training yourself at the same time by reading aloud the user manual. This allows your computer more time to get to know your voice and the way you pronounce things.

I find the whole concept fascinating. Usually, my thoughts appear on the page as I'm thinking them, but with speech-to-text I find myself pausing a lot to consider how my words sound before saying them. I think the type of book I would wright using speech-to-text might be very different.

If that were the only delay, perhaps I'd persist, but there are other issues.

I tried an opener:

The night was cold. Snowflakes fell like falling hope between the cobblestones. White winter swept the town, holding the lives of young and old between its skeletal fingers.  

and got:

Of the the the night was cold.  Snowflakes fell like falling home between the cobblestones.  White winter swept the town holding the lives of young and old between their skeletal fingers.

Perhaps it's an old microphone, but when I was silent for too long, thinking how to begin, the programme filled the silence for me with a couple of extra thes.

I threw it the old tongue twister Peter Piper picked a peck of pickled peppers and received Peter Piper picked up at him for a month for.

Perhaps this would improve with time, but the process of correction is tiring. Each time there is a mistake you have to tell the program correct [word] and either choose a number from the list:


Or spell it out letter by letter, correcting each letter it mishears.


That's really the problem. In the time it takes to correct the mistakes in one sentence, you could have written twenty more. Speech-to-text has still got a long way to go. Which might be why it's tucked away right at the very back of the Windows tool box. I've heard Dragon is pretty impressive, but it's pricey and you'd want to test it out before purchasing.

No comments:

Post a Comment