BUILDING AN AUTOMATIC ANSWERING MACHINE TO PANEL SURVEY QUESTIONS

There are now problems interviewing people by phone, even for simple and short interviews. This for three main reasons: 1.Inflation of commercial and/or promotional interviews 2.The decrease in the number of landline phones 3.Connecting people via mobile phone due to the absence of public directories of telephone numbers is extremely expensive.

The SVTP project aims to reduce these difficulties by the construction of an answering machine. The machine will learn to answer from previous experiences, from surveys already conducted on subjects of interest (tourism in our case). The project we will work on a panel of tourism surveys to create a virtual survey campaign by means of the automated answering machine. The proposed machine will be able to answer survey questions. We have already done 10 surveys in years 20,21,22,23 about tourism in Piedmont for the local administration. Three for Sommers, three for Autumns, three for Winter and one for testing purpose (Autumn 23). We have already built a base line automated survey from the previous nine surveys and compared the answers of this survey on the test survey (Autumn 23) which results will be presented. Furthermore, several simulation trials using different forecasting techniques will be done: logistic regression and decisional trees for example. The models test will compare the model results with the baseline model to better preview the number of future Piedmont tourist visitors. The results of the test will be presented. Something like what ChatGPT does building artificial questionnaires, just to give an example. We will generate the answers instead. Philosophic implications of all of this are obvious. I will give you some clues in the presentation.

Turin, 12th January 2224