More survey questions.

[evanl@peopleclues.com]

Hi there, couple more questions for you on the Survey.

1. How are the user inputs stored? Do you guys store each answer as the user go along or do you wait until the survey is done? What happens to the answers if the user hangs up half way through? Do you still store the partial answers in the system? If you do, would we be able to get access to these partials if we need to?

2. When we use a full name question type, what does your system store? Do you store the sound file of the user say and spelling their name or do you just keep the speech to text value of their name? If you do store a voice file for it, would we be able to get to it?

Thanks!

[support]

1. The IVR platform stores the answer to each question immediately. The default setting is to discard the whole IVR session if the user doesn't complete the IVR survey, however there is a very simple toggle option in Survey Settings that allows you to store these partially completed sessions, e.g., if the user hangs up halfway through an IVR survey. In that case, every answer the phone system was able to collect from the user before the call ended will be saved to our database, and the data will show up in your reports and individual responses along with everything else. So the simple answer is yes, we can store and provide you the data for partial responses.

2. For full name questions, if you have speech recognition enabled, we will first try to use speech recognition to collect the first and last name. Our recognition grammar covers about 90% of American English first and last names. If this succeeds, you will get a text output of what our IVR grammar engine recognized. If this fails more than twice, or if you choose to disable speech recognition in Survey Settings, the user will be prompted to record their name. These recordings are fully available to you in the individual responses section and for batch downloading.

Hope this helps!

[evanl@peopleclues.com]

When the person does not successfully get their name recognized and have to record their name in a voice file, what do you send back to as the value for that question?

I currently have
QUESTION_TEXTS_0 What is your name.

When I called in and the system asked me to record my name since it didn't recognize it, it ended up sending this as the value:
ANSWERS_0 C:\CFusionMX7\runtime\servers\coldfusion\SERVER-INF\temp\wwwroot-tmp\neotmp38937.tmp
It appears to not send back just a null value of some sort.

Second question. It doesn't seem like the speech to text engine recognize hyphenated names. I spelled out a name with a hyphen in it, and it did not recognize it. What does the user need to say for the system to recognize a hyphenated name?

Thanks!

[support]

The documentation on IVR subdialogs for IVR surveys contains the information on how audio data is passed back to you. Specifically:

The encoding type for the POST request will be multipart/form-data, because some of the answers may consist of recorded audio files with a Internet media type of audio/basic; these are 8 kHz, 8-bit mu-law-encoded headerless WAV files that the web server hosting your subdialog may choose to save and process.

We aren't sure how your ColdFusion server is set up but "C:\CFusionMX7\runtime\servers\coldfusion\SERVER-INF\temp\wwwroot-tmp\neotmp38937.tmp" may refer to the filename of the temporarily saved file on your disk. Note that this is not a normal WAV file because it is headerless--you will have to convert it using a utility like SOX if you want to play it with a media player later. This IVR thread has an example of the exact command you can use.

As for hyphenated names, our IVR grammar for full names ignores hyphens, so the user can spell it without saying anything with regard to the hyphen and it will recognize common names like so (e.g., Jean-Marie becomes Jeanmarie). Unfortunately we don't yet have data on all the hyphenation variations for common names.

[evanl@peopleclues.com]

So, what you're saying is that if you're not returning the speech to text recognized string, you're returning the .wav file as part of the submission?

[support]

Yes, it is POST'ed back as a multipart attachment.

Like IVR developers warned though, it is not a regular ".wav file", it is mu-law and headerless; you need to use a audio conversion utility like SOX to convert it to a typical WAV file with a header, or whatever other format you prefer.