Semiautomatic dictionary-based tweet classification for measuring well-being

In this paper we describe a semiautomatic dictionary-based approach to filter tweets talking about specific topics. In particular, we are interested in studying the citizen well-being (WB) and, for this aim, we select tweets pertaining two WB dimensions such as environment and health. For this purpose, we use dictionaries containing keywords selected by analyzing tweets published by some Official Social Accounts linked with the two topics. The selected tweets are then processed in order to estimate the sentiment of the population with respect to such specific subjects. In this paper, we present some preliminary results for Great Britain (GB) using tweet collected on the whole country for the six-weeks period from 2019/01/14 to 2019/02/24. The results show that, on the one hand, our dictionary-based classification approach reaches good levels of accuracy, sensitivity and specificity; on the other hand, we assess the spatial variability across GB of the two dimensions we are studying by means of the tweets sentiment analysis.