Over the last decades, and especially since the growing use of smart mobile devices, there has been a significant increase in the dissemination of geographic information. In this context, the term ‘Volunteered Geographic Information (VGI)’ describes the widespread engagement of citizens in the production of spatial information. The trend of VGI has been strongly fostered by the popularisation of social networking services with Twitter currently being the fastest growing online network.
By analysing geo-tagged tweets of individual users, new information can be retrieved and put into a wider context, allowing for the investigation of particular patterns in space. This study aimed at elaborating a method to visualise specific Twitter content. For this purpose, the mood of tweeters in the United Kingdom (UK) was analysed. Tweets from February 2012 were queried for ‘I love’ and ‘I hate’ keywords resulting in 49917 data points that were processed and visualised.
Workflow
The pre-processing of the Twitter data was undertaken in PostgreSQL. For performance reasons, the table was indexed, and the tweets were filtered for ‘I love’ and ‘I hate’ keywords using the Structured Query Language (SQL). To georeference the data, the geolocation code of each tweet was transformed from well-known-text format (WKT) into x and y coordinates, using ArcGIS. For the presentation of the data, a kernel density estimation was applied, calculating the density surface of the tweet distribution across the map area. In addition, the raster calculator was used, computing the share of ‘love’ and resp. ‘hate’ tweets on all messages. As it was found that the ratio of ‘I love’ to ‘I hate’ Tweets was approx 3:1, this mean value was subtracted from the ratio surface. The map visualises the deviation from the mean love to hate ratio.
