LEARNING TO RECOGNIZE MISSING E-MAIL ATTACHMENTS

摘要

Forgotten attachments of E-mail messages are a common and obnoxious problem. Several E-mail readers provide plug-ins that attempt to tackle this problem by trying to guess whether a message needs an attachment and warn the user in case she or he does not attach a file to such a message. However, these approaches essentially only work with a fixed list of keywords, which trigger such a warning whenever they occur in a message. In this article, we try conventional machine learning techniques, which have been previously shown to work well for related problems such as spam mail filtering, on this new problem. Our results show that they work very well, clearly outperforming simple keyword-based approaches. The software is available as plug-in for the Thunderbird E-mail reader.