The widespread adoption of highly interactive social media like Twitter, Facebook and other platforms allow users to communicate moods and opinions to their social network. Those platforms represent an unprecedented source of information about human habits and socio-economic interactions. Several new studies have started to exploit the potential of these big data as fingerprints of economic and social interactions. The present analysis aims at exploring the informative power of indicators derived from social media activity, with the aim to trace some preliminary guidelines to investigate the eventual correspondence between social media indices and available labour market indicators at a territorial level. The study is based on a large dataset of about 4 million Italian-language tweets collected from October 2014 to December 2015, filtered by a set of specific keywords related to the labour market. With techniques from machine learning and user's geolocalization, we were able to subset the tweets on specific topics in all Italian provinces. The corpus of tweets is then analyzed with linguistic tools and hierarchical clustering analysis. A comparison with traditional economic indicators suggests a strong need for further cleaning procedures, which are then developed in detail. As data from social networks are easy to obtain, this represents a very first attempt to evaluate their informative power in the Italian context, which is of potentially high importance in economic and social research.
Tweet-tales: moods of socio-economic crisis?
Grazia Biorci;Antonella Emina;Lisa Sella;
2016
Abstract
The widespread adoption of highly interactive social media like Twitter, Facebook and other platforms allow users to communicate moods and opinions to their social network. Those platforms represent an unprecedented source of information about human habits and socio-economic interactions. Several new studies have started to exploit the potential of these big data as fingerprints of economic and social interactions. The present analysis aims at exploring the informative power of indicators derived from social media activity, with the aim to trace some preliminary guidelines to investigate the eventual correspondence between social media indices and available labour market indicators at a territorial level. The study is based on a large dataset of about 4 million Italian-language tweets collected from October 2014 to December 2015, filtered by a set of specific keywords related to the labour market. With techniques from machine learning and user's geolocalization, we were able to subset the tweets on specific topics in all Italian provinces. The corpus of tweets is then analyzed with linguistic tools and hierarchical clustering analysis. A comparison with traditional economic indicators suggests a strong need for further cleaning procedures, which are then developed in detail. As data from social networks are easy to obtain, this represents a very first attempt to evaluate their informative power in the Italian context, which is of potentially high importance in economic and social research.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.