Scores of feature words and earthquake-related words for each cluster before and after Japanese earthquake.

<p>We analyse tweets that are collected in the interval of one week before and after 11th March 2011, when huge earthquakes occurred in Japan. (a)–(d) Average scores of the feature words shown in <a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0139085#pone.0139085.g001" target="_blank">Fig 1</a>. All words shown as feature words in <a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0139085#pone.0139085.g001" target="_blank">Fig 1</a> represent characteristics of each cluster well, since all the average scores of the feature words are greater than the others. (e)–(h) Scores of some earthquake-related words within each clusters introduced in <a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0139085#pone.0139085.g001" target="_blank">Fig 1</a>. (e) Everyone mentions earthquake-related words, possibly because the earthquake was too big for any group of people to ignore. (f), (g), and (h) Some words have a spike mainly in affiliated clusters. For instance, the word <i>shake</i> has a spike only in ‘English speaking’. Details are discussed in the main text.