User Study

Abstract

In Chapters 3 and 4.6 we introduced new methods to address very common message addressing problems, namely email recipient recommendation and email leak prediction. Although the proposed methods showed promising results in batch experiments on very large email collections, many questions were still unanswered. How can these methods be incorporated in an integrated interface? Can users notice any difference in quality between rankings provided by different baseline algorithms? Can these methods really catch email leaks? Can we estimate how often email leaks occur? Can these techniques be adopted and benefit a large number of email users?