Improve Tags Recognition with a Simple Fix

I love the tags feature (“Show Tags”, followed by the tag e.g. “CB” or “D3”)

The problem is that the two letter tag is often recognized inaccurately. This is a big problem (for me at least) on a complex interface such as Gmail (where over 100 tags appear). For example,

  • C3 might be recognized as CC
  • D3 as PP
  • GS as CS

Anyway, I’ve seen a quick and simple fix for this problem in other voice recognition software. And, it’s nearly bulletproof. Basically, instead of using two characters for tags, use a two digit number. Tags like “eleven”, or “thirty three” or “sixty five” have much higher recognition accuracy than two character tags.

Can you make this change? Or, at least consider it as an option?


Another possible solution (though not quite as good, IMO) would be to continue to use alpha-numeric tags but instead also accept the NATO Phonetic Alphabet for letters (alpha for A, bravo for B, charlie for C, etc.).

This approach has the downside that it requires us users to learn a new alpha, but we might do it… because the tags recognition errors are so annoying.


Welcome Yu-Shen Ng! Great feedback!

We tried to eliminate all combinations that produced inconsistent results, but of course results may vary among speakers. Since enough people have requested it, I will revamp the tags names in a subsequent version - using more digits and less letters.

Letters can be thought of a as a power user feature. They are faster because of less syllables (eg. most 2 digit tags will be at least 3 syllables) and they give more possibilities with just two characters compared to digits.

I think we can find a better in between. I like your idea of allowing a radio alphabet. We have a similar feature for the spell commands - but instead of restricting it to the radio alphabet, users can use any word that starts with the letter they need.

Another thing we can do is allow customizing the tags list. This would bring in interesting possibilities.

Custom radio alphabets for tags would be cool. I’ve been using Talon recently and been quite impressed, the selection of words for the alphabet works really well. (scroll down here for the list of 26 words if you haven’t seen it)

seconding this feature! i’ve had a lot of problems with accuracy in tags, especially with letters like n and m. i often can’t get lip surf to recognize the correct tag, even if i say it repeatedly. i’d rather have more syllables to say if it meant more accuracy