Estimation of User Location and Local Topics Based on Geo-tagged Text Data on Social Media

This paper proposes a method for estimating microblogging user location to determine local topics of importance based on area-specific term cooccurrence. Geotagged information on social media has not previously been sufficient to determine local topics, however, the amount of information available on social media has continued to expand due to the widespread use of smartphones. Notably, the amount of information generated from regional cities is significantly smaller than that from metropolitan cities. Hence, we must estimate the location of each user in a regional city to obtain adequate local information for determining local topics. To extract this information, we define area-specific scores of terms and co occurrences that are calculated using term frequency, as well as average and standard deviation of the longitude and latitude of raw geotagged information.