|
Abstract:
|
The world is engulfed in a polarizing environment due to individuals who hold firm opposing opinions. Recently, inciting, large-scale events and increased news coverage has enlightened this subject. An amalgamation of these opposing individuals, large-scale events, and news coverage creates a hyper-polarized community. Hyper-polarization can lead to violent acts, so we address this issue by developing a system of hierarchical models to model news coverage through news articles and their sentiment across time. Specifically, we apply topic modeling techniques to detect essential topics per day and match topics across time, which we match implementing a ranked, non-parametric variant of correlation. We quantify the influence of these temporally varying topics by using a high-dimensional multilevel version of a system of Dynamic Linear Models. With the combined implementation of topic models and time varying models, the proposed method accounts for both time and topic dependence. In conclusion, we identify underlying associations between topic polarity and known highly polarizing events.
|
