A1380
Title: Converting textual data into structured survey data: A ChatGPT approach
Authors: Giancarlo Manzi - University of Milan (Italy) [presenting]
Aurea Grane Chavez - Universidad Carlos III de Madrid (Spain)
Qi Guo - Universidad Carlos III de Madrid (Spain)
Abstract: A new approach to transforming textual data into survey data with the use of chatbot technologies is presented. ChatGPT APIs are considered to associate each respondent's response to an open-ended question (included in a 2019 survey questionnaire about a bike-sharing service in Milan, Italy) to the most probable Likert scale question in the questionnaire. ChatGPT is also asked to give, according to the meaning of each of such responses, a possible rating answer on a Likert scale for the chosen question. In this way, the congruence of the answers is evaluated to the open-ended question to the Likert scale options chosen by respondents and form a reliability measure to be compared to standard questionnaire reliability measure s like test-retest reliability, inter-rate reliability, etc.
