Where's @Waldo?: Finding Users on Twitter

In today's social media world we are provided with an impressive amount of data about users and their societal interactions. This offers computer scientists among others many new opportunities for research exploration. Arguably, one of the most interesting areas of work is that of predicting events and developments based on social media data and trends. We have recently seen this happen in many areas including politics, finance, entertainment, market demands, health, and many others. Furthermore, there has been a lot of attention garnered on being able to predict a user's location based on their online activity taking into account that large amount of social interaction online is done behind usernames and anonymous titles. This area of research is well-known as geolocation inference. In this paper, we propose a novel model for geolocation inference of social media users using the aid of a discrete event: the Solar Eclipse of 2017. Being able to use the path pf the eclipse and timing of its path of travel to infer a user's location is a unique model seen only in this paper. We apply this unique model to Twitter data gathered from users during the Solar Eclipse of 2017 and attempt to determine if certain features of the data itself are indicative of users viewing the eclipse or of similar events. Taking advantage of Stanford's natural language processing software, we also consider the proportions and existences of many words, part-of-speech tags, and relations between users both found in our sample data, in an attempt to find key features of users who are viewing the eclipse. We discuss our results using our unique model and conclude by discussing the strengths and weaknesses of the model with the resulting potential future work.