Early Tracking of People's Reaction in Twitter for Fast Reporting of Damages in the Mercalli Scale

摘要

The Modified Mercalli Intensity Scale is a measure of the severity of an earthquake for a nonscientist. Since the Mercalli scale is based on perceived effects, it has a strong dependence on observers. Typically, these reports take time to be prepared and, as a consequence, Mercalli intensities are published hours after the occurrence of an earth-quake. The National Seismological Center of Chile needs to provide a preliminary overview of the observed effects of an earthquake. This has motivated us to create a system for early tracking of people's reaction in social networks to infer Mercalli intensities. By tracking people's comments about the effects of an earthquake, a collection of Mercalli point estimates is retrieved at county level of granularity. We introduce the concept of Reinforced Mercalli support that combines Mercalli point estimates with social support, allowing to discard social unsupported estimates. Experimental results show that our proposal is accurate providing early Mercalli reports 30 min after an earthquake, detecting the maximum Mercalli intensity of an event with high accuracy in terms of mean absolute error (MAE).