| 摘要 |
A classification unit (120) puts mails into one of a plurality of categories according to a classification rule. A learning-sample mail extraction unit (140) reclassifies the mails into any one of a plurality of categories according to a learning sample extraction condition (170). The learning-sample mail extraction unit (140) extracts, as mails for the learning samples, only the mails in which the category classified by the classification unit (120) coincides with the category reclassified according to the learning sample extraction condition (170). A learning unit (150) learns mechanically using the mails for the learning sample extracted by the learning-sample mail extraction unit (140), and updates the classification rule. |
