On the Characterisation and Clustering of the Blogosphere

s – C02 13.41% ± 0.26 15.89% ± 0.26 13.89% ± 0.32 10.03% ± 0.27 8.64% ± 0.28 5.28% ± 0.31 6.99% ± 0.26 7.09% ± 0.32 18.78% Short news –R8 17.71% ± 0.04 14.89% ± 0.01 17.22% ± 0.02 13.56% ± 0.02 8.52% ± 0.02 7.33% ± 0.01 7.04% ± 0.01 5.31% ± 0.01 8.42% Weblogs – WbN 19.98% ± 0.05 18.20% ± 0.03 18.69% ± 0.02 16.84% ± 0.02 8.25% ± 0.01 6.23% ± 0.01 5.01% ± 0.01 3.16% ± 0.01 3.64% Weblogs – WbW 18.69% ± 0.01 18.91% ± 0.01 20.51% ± 0.01 17.67% ± 0.01 8.6% ± 0.01 5.85% ± 0.005 4.28% ± 0.006 2.37% ± 0.005 3.12% Weblogs – WbLJ 22.40% ± 0.03 21.60% ± 0.04 19.60% ± 0.03 14.47% ± 0.02 6.47% ± 0.02 4.37% ± 0.01 4.08% ± 0.01 2.12% ± 0.01 4.89% Microblogs – Mb 26.31% ± 0.04 13.40% ± 0.02 16.19% ± 0.02 13.83% ± 0.01 8.66% ± 0.02 7.73% ± 0.01 5.4% ± 0.01 3.22% ± 0.01 5.26% Microblogs and weblogs showed consistent percentages i.e., not big variations over the lengths less than or equal to four, and as expected we obtained high percentages for short tokens (from one to four characters) due, among other factors, to slang or shortening words and informal text (Yang, 2011). Based on this information we can say that if more than 50% of a corpus contains tokens of length less or equal to four, it is highly probable to be a corpus of microblogs.