Identifying sound recording to the questions

Discussions about CSEntry
Post Reply
AriSilva
Posts: 620
Joined: July 22nd, 2016, 3:55 pm

Identifying sound recording to the questions

Post by AriSilva »

Hy folks,
We are using sound recording to keep track of what is going on in the field, if the interviewers are asking correctly the questions, if they are not answwering them by themselves, etc.
It is working well, we discovered a buch of bad interviewers, but...
It is difficult to identify, by hearing the recording, which questions are being answered, mainly because we have a very lengthy questionnaire, with several blocks. We even split the sound recording by questionnaire blocks, each block is recorded to a different file, but even doing that sometimes the thing is not clear.
Is there a way to connect a specfic part of the sounding to a specific question, that is a way to keep track of where the interviewer is in the questionnaire?
Best
Ari
Best
Ari
Gregory Martin
Posts: 1876
Joined: December 5th, 2011, 11:27 pm
Location: Washington, DC

Re: Identifying sound recording to the questions

Post by Gregory Martin »

Using CSPro functionality that currently exists, what I would do is:

1) Enable paradata logging for your application.

2) By default, paradata logging keeps track of when a field is entered and when the operator leaves the field. It contains timestamps (UNIX time) for all of paradata events.

3) When you start the background recording, use the logtext function to note, in the paradata log, that you started the recording: https://www.csprousers.org/help/CSPro/l ... ction.html

4) Finally, with some queries in the paradata log, which is a SQLite file, you can find the logtext entry that gives you the timestamp of when the audio recording started. You can then subtract this value from the field entry/exit values. The resulting value is the number of seconds into your recording that the field was entered/exited.

This certainly isn't trivial, but if you did this in a script, you could write code to do this once and then output information about all of your recordings.
Gregory Martin
Posts: 1876
Joined: December 5th, 2011, 11:27 pm
Location: Washington, DC

Re: Identifying sound recording to the questions

Post by Gregory Martin »

Another option that may actually be easier is to use the undocumented global On_Focus function. This is like the PROC onfocus event, but it is called for all procedures. You can use this to write information about what field is being entered using the getsymbol function. For example:
Audio recording;
File audioTimesFiles;
numeric audioStartTime = notappl;

function StartRecording()
    if recording.record() then
       
audioStartTime = timestamp();
        audioTimesFiles.open(maketext("audio-recording-times-%d.csv", audioStartTime), create);
    endif;
end;

function StopRecording()
    recording.stop();
    audioTimesFiles.close();
    audioStartTime = notappl;
end;


function On_Focus()
    if audioStartTime <> notappl then
       
audioTimesFiles.write("%f, %s", timestamp() - audioStartTime, getsymbol());
    endif;
end;
AriSilva
Posts: 620
Joined: July 22nd, 2016, 3:55 pm

Re: Identifying sound recording to the questions

Post by AriSilva »

Thanks very much Greg, I´ll try them.
Best
Ari
Best
Ari
Post Reply