PROBLEM TO BE SOLVED: To calculate a feature capable of appropriately discriminating a spam page even if the spam page is automatically generated.;SOLUTION: A spam account score calculation device for calculating a score showing spam likelihood of an account name comprises: account language model generation means which calculates a frequency in spam and a frequency in non-spam for each token obtained by decomposing the account name, from aggregation of spam account names and non-spam account names, and stores it in an account language model DB as an account language model; and account evaluation means which calculates probability of spam and non-spam for each token in an estimation target account name on the basis of the account language model stored in the account language model DB and calculates a score showing spam likelihood of the estimation target account name on the basis of the probability.;COPYRIGHT: (C)2014,JPO&INPIT
展开▼
