O-025 n-gramモデルを用いた日本語スパムフィルタの設計及び構築(O.情報システム)

概要

論文の詳細を見る
Nowadays many people feel uncomfortable because of receiving unnecessary mails that is often called spam or junk mails. Worse still, spam mails trend to increase year by year. So automatically rejecting spam mails, that is, spam filter has been researched recent years. If a mail contains some words which appear in spam mails received before, and if it does not contain other words which rarely appear in spam ones, then there is a high probability that the mail may be a spam one. Spam judgment for a current mail can be performed by calculating conditional probability on the basis of the database constructed with a lot of mails received in the past. This is a Bayesian spam filtering method. The authors propose an improved Bayesian spam filter. The n-gram model built in its filter is suitable for processing Japanese language in order to detect a Japanese spam mail. It is also possible to detect a spam mail written in multi-byte languages such as Chinese and/or Korean. The proposed Bayesian spam filter has been developed in the programming language Ruby, which is an object-oriented script language.
FIT(電子情報通信学会・情報処理学会)推進委員会の論文
2004-08-20