Context-Aware Systems and Applications. 5th International Conference, ICCASA 2016, Thu Dau Mot, Vietnam, November 24-25, 2016, Proceedings

Research Article

Personalized Email User Action Prediction Based on SpamAssassin

Download
355 downloads
  • @INPROCEEDINGS{10.1007/978-3-319-56357-2_17,
        author={Ha-Nguyen Thanh and Quan-Dang Dinh and Quang Anh-Tran},
        title={Personalized Email User Action Prediction Based  on SpamAssassin},
        proceedings={Context-Aware Systems and Applications. 5th International Conference, ICCASA 2016, Thu Dau Mot, Vietnam, November 24-25, 2016, Proceedings},
        proceedings_a={ICCASA},
        year={2017},
        month={6},
        keywords={Personalized email prioritization Email user action SpamAssassin},
        doi={10.1007/978-3-319-56357-2_17}
    }
    
  • Ha-Nguyen Thanh
    Quan-Dang Dinh
    Quang Anh-Tran
    Year: 2017
    Personalized Email User Action Prediction Based on SpamAssassin
    ICCASA
    Springer
    DOI: 10.1007/978-3-319-56357-2_17
Ha-Nguyen Thanh1,*, Quan-Dang Dinh2,*, Quang Anh-Tran3,*
  • 1: Hanoi Department of Information and Communications
  • 2: Hanoi University
  • 3: Posts and Telecommunications Institute of Technology Hanoi
*Contact email: nguyenthanhha_sotttt@hanoi.gov.vn, quandd@hanu.edu.vn, tqanh@ptit.edu.vn

Abstract

Email overload, even after spam filtering, causes waste of time and reduction of work efficiency to email users. Email prioritization is the general solution for the problem. The idea is to sort incoming emails in a decreasing order of importance so that the most important messages are read and processed first and less significant ones later, if there is enough time. This paper proposed a method to predict the action that a user would take on an email. The method is based on SpamAssassin, a famous spam filter framework. Instead of classifying emails as spam and ham (non-spam message), this method is used to predict amongst the three most common actions: reply, read and delete. Experiments are conducted to measure the effectiveness of the new method on a dataset built by the authors.