Principle Component Analysis of Twitter Timeseries

In order to facilitate a better understanding of the kinds of temporial activity we ran PCA on all 1-grams, hashtags, and handles ever to break the top 1000 words on Twitter in any langauge.

On the right side we see perhaps three large arms of points which are consistantly used, as compared to some of the more circular clusters on the left which are composed of words which spiked on a few days, or had huge spikes but then decayed quickly. The ring-like structure surrounding the cluster in the lower left, correspondes to exogenous events with decay back into the lexical abyss, with the time of the spike moving back in time as one moves counter-clockwise. Many events in this ring correspond to the usage of dates in the format %y%m%d, or 180908 for September 8th 2018.