Classificazione manuale vs automatica dei resoconti delle chiamate ricevute da Telefono Amico Italia (2016 -2020)

Telefono Amico Italia (TAI) is a voluntary organization that provides a telephone help service for people in emotional distress. The telephone operators summarize in short reports the contents of calls and a group of experts classifies them into categories. The present contribution moves from the analysis of the contents of the reports (203.316 in 2016-2020) with a twofold aim of detailing the categories adopted by TAI and proposing an update. Two different automatic classification perspectives have been exploited: a top-down one which aims to assess the consistency of the contents of the phone calls with the categories adopted by TAI's experts, and a bottom-up one which seeks to obtain a classification of the texts by topic (topic extraction). By means of statistical analysessuni and multivariate analyses, the categories that emerged from the classifications were observed in relation to the characteristics of the appellants. The top-down approach, showed the goodness of the categories adopted by the experts, but also the need for a redefinition of some categories that showed large areas of semantic overlap. The bottom-up analysis, has both confirmed the presence of categories already adopted and allowed to recognize the existence of new emerging categories. The multivariate analysis of these new categories revealed a decisive role of the category mental health in monitoring suicide.