Speech-to-text may never be good enough

The first day of this new year is a good time to look back at last year and think about lessons learned. Those following my GoFundMe campaign are aware that 2016 was not a good year for me.

2016 started with a broken arm that sidelined me for a few weeks. As you can gather from the X-ray at right, this required surgery to set the bones and a few weeks with the arm in a sling. This affected my ability to write.

It didn’t help that I am right-handed and the broken arm was my right one. Soon the inability to write was driving me crazy.

After careful thought I wondered if speech-to-text technology had improved over the years to make it the solution to my problem. I set out testing it on iOS, Android, and Windows to find out.

What I discovered was that all of these platforms with voice dictation were about the same. And that wasn’t very good. Input by voice has been tweaked to make it good for searches but not dictation like I needed.

Understanding what I was speaking would usually start off OK but then it would go so very bad. A short article that would usually take me half an hour would often take hours with all the proofreading I was doing just to fix the erroneous speech conversion. Once the system gets off it gets very far off since they al use context to figure out what you meant.

Fortunately my fingers were exposed by the sling and I ended up tapping by writing into a keyboard. This was even faster than voice input and that says a lot about where voice input technology is today.

Leave a Reply

Your email address will not be published. Required fields are marked *