(UPDATE 3/14/09: take a look at the 10 ways to optimize speech recognition quick guide.) 

There have been lots of posts all over the web about how cool Response Point’s speech recognition is.  But every now and then somebody tells us it’s not working out so well for them.  If you know somebody who’s in this situation, here are some tips to try:

  1. Make sure the name entered in the system is something that people actually say.  There are a few variants of this:
    1. If your name is entered as “Robert”, but nobody ever calls you that, make sure you enter an appropriate nick name (e.g. “Rob”), or whatever name people actually call you by.
    2. Make sure your first name is in the first name field, and your last name is in the last name field.  Sounds obvious, but worth checking.  Nobody ever calls me “mister Robert”, so “Robert” shouldn’t be in my last name field.
    3. Don’t use names like “Sales Manager Robert”.  Nobody’s ever going to say “Sales Manager Robert”.  Instead, try something like entering first name = “Robert”, last name = “Brown”, and nick name = “Sales Manager”.  Or create a “sales” group extension and put it on the same phone (see my ninja skills post for that one…).
    4. Sometimes you may have people in the system who have very similar names.  For example, “Robert Brown Jr.”, and “Robert Brown Sr.”  If a caller asks for “Robert”, the system will have no idea which one is wanted, and ask the caller to choose.  This is a hassle for the caller.  So make sure you use nick names to distinguish them.  e.g. “Junior” and “Senior”, or “Rob Junior” and “Rob Senior”, or however you tell these guys apart in the real world. (“Luke” and “Darth”?)
  2. The system can also respond poorly to unnatural speech patterns.  For example:
    1. Yes, computers are definitely idiots, but they’re fast idiots, and … if … you … speak … unnaturally… slowly… the system might just confuse one of your pauses for the end of your statement, and carry on as if you were actually finished talking.  Speak at a normal pace, at the same speed you would speak to a real person.
    2. If your voice is very loud, the sound signal can actually be clipped by the phone, which makes speech recognition less accurate.  No, I’m not saying you’re loud.  You’re way cool.  But computers and electronics don’t know that.  Their microphones are tuned for a certain range of volumes and frequencies.  So if you have a louder than average voice, see if speaking a tad softer makes a difference.
    3. Similar thing if you have a very quiet voice.  Especially if you’re in a noisy environment.
    4. Shorter names are harder to recognize, especially if there are a lot of short names in the system. Sometimes it helps to say somebody’s full name.