An AI-based approach for transforming textual information into structured survey data

In this work we present an original approach for transforming textual data in survey data with the use of chatbot technologies. We consider ChatGPT APIs to associate any short text (regarded as a ”possible answer”) to the most probable question which could generate it among a list of few questions, possibly representing a survey questionnaire. We also ask ChatGPT to give, according to the meaning of each short text, a possible answer option. This short text might be, for example, (i) an X tweet or (ii) an open-ended question in a real survey questionnaire. In case (i) we construct a survey data set from X messages, and in case (ii) we use this procedure to check for answer reliability in a real survey.