IVR Deconstructed

Just wanted to get some tips out there to IVR developers who are designing applications and have to factor in noisy environments.

1) Set specific local properties for when you are prompting the user for an input.

So, if you had a <field> where you are prompting the user to say their name, you may want to set specific values for properties that work with speech recognition such as: sensitivity, incompletetimeout, and confidencelevel

2) Set a low sensitivity value if you believe background noise (i.e. city traffic, construction noise) will be a factor in your IVR application.

Setting a low sensitivity will help capture inputs when the caller is in a noisy environment. Typically, a value of 0.3 or 0.4 should suffice in this situation.

<property name=”sensitivity” value=”0.4″/>

3) Increasing the incompletetimeout value within your <field>.

Setting a higher incompletetimeout value will give your caller more time to input a response before the speech recognizer recognizes the IVR response. This particularly helps when the caller is saying a long string of digits and pauses in between saying digits.

<property name=”incompletetimeout” value=”3s”/>

4) Setting prompt bargein=”false”

By setting prompt bargein=”false” within your application, you can prevent the caller from accidentally barging in with an incorrect input by coughing or with a loud background noise.

<prompt bargein=”false”>
Please say your nine digit social security number.
</prompt>

Hope these tips help you with designing your IVR application.