Using Twitter to Better Understand the Spatiotemporal Patterns of Public Sentiment: A Case Study in Massachusetts, USA.

Abstract

Twitter provides a rich database of spatiotemporal information about users who broadcast their real-time opinions, sentiment, and activities. In this paper, we sought to investigate the holistic influence of land use and time period on public sentiment. A total of 880,937 tweets posted by 26,060 active users were collected across Massachusetts (MA), USA, through 31 November 2012 to 3 June 2013. The IBM Watson Alchemy API (application program interface) was employed to quantify the sentiment scores conveyed by tweets on a large scale. Then we statistically analyzed the sentiment scores across different spaces and times. A multivariate linear mixed-effects model was used to quantify the fixed effects of land use and the time period on the variations in sentiment scores, considering the clustering effect of users. The results exposed clear spatiotemporal patterns of users’ sentiment. Higher sentiment scores were mainly observed in the commercial and public areas, during the noon/evening and on weekends. Our findings suggest that social media outputs can be used to better understand the spatial and temporal patterns of public happiness and well-being in cities and regions.