Social networks VGI: Twitter sentiment analysis of social hotspots

Abstract

The enormous amount of data generated on social media provides vast quantities of geo-referenced data. Volunteered Geographic Information (VGI) originating from social networks has produced new challenges for research and has opened opportunities for a wide range of use cases. Smartphones with built-in GPS sensors enabled users to easily share their location and with the growing number of such devices available, VGI data is expanding at a rapid rate. Twitter is one of the most popular microblogging services. It’s a social network that enables access to the data that is being created on the platform. It also allows for real-time retrieval of data from a given geographic area. In this paper we give an overview of a system for detecting and identifying social hotspots from Twitter stream data and applying sentiment analysis on the data. Utilizing the Twitter Streaming Application Programming Interface (API), we collected a significant number of Tweets from New York and we evaluated the quality of the retrieved data. In this paper, we outline advantages and disadvantages of using various clustering algorithms over the data for this purpose, namely hierarchical agglomerative clustering and DBSCAN. We also elaborate on techniques for identifying social hotspots from spatially localized clusters. Finally, we present a deep learning approach to sentiment analysis used to determine the attitude of users participating in the identified social hotspots.