Centering the researcher in big data: a topical analysis of Covid-19 tweets

Keywords: digital geography, big data, discourse analysis, sentiment analysis, covid-19
Abstract Type: Paper Abstract

Abstract

Data is simply too “big” for any one person to examine in its entirety. As a result, researchers rely on machine learning (ML) or artificial intelligent (AI) programs to handle, analyze, and derive meaningful insight. While researchers interact with the inputs and outputs of the programs, the internal state is shrouded in mystery. Rather than being at the center of the analysis, researchers move to the periphery. This project proposes the sentiment-cluster-topic extraction model to not only assist researchers in mining big data, but also to bring researchers back into the center of it. Using Twitter data on Covid-19 tweets, this project asks: how can we classify and organize tweets –how can we capture the overall discussion and ideas expressed within a particular cluster, e.g., what topics are most commonly discussed or articulated? And how we can do this while keeping the researcher central to this process? This project uses three algorithms to provide researchers the means to investigate their data: sentiment analysis (VADER algorithm), cluster analysis (K-means), and topic analysis (LDA). While the algorithms organize and classify the data, the point of this model is that it provides researchers a point of departure for analyzing and understanding their data, researchers to investigate any cluster of data and repeat the analysis on it to derive more gradual insight.