In this paper, we propose a methodology quantifying temporal patterns of nonlinear hashtag time series. Our approach is based on an analogy between neuron spikes and hashtag diffusion. We adopt the local variation, originally developed to analyze local time delays in neuron spike trains. We show that the local variation successfully characterizes nonlinear features of hashtag spike trains such as burstiness and regularity. We apply this understanding in an extreme social event and are able to observe temporal evaluation of online collective attention of Twitter users to that event.