mine social network person email priorit shinja yoo yime yang frank lin il chul moon email preval commun tool todai solv email overload problem pressingli urgent good allevi email overload automat priorit receiv messag prioriti user research statist learn method fulli person email priorit pep spars due privaci issu peopl reluct share person messag import judgment research commun import develop evalu pep method assumpt limit train exampl system person email data user train test model user paper present studi knowledg assumpt specif focu analysi person social network captur user group obtain rich featur repres social role viewpoint user develop semi supervis transduct learn algorithm propag import label train exampl test exampl messag user node person email network method enabl obtain enrich vector represent email messag consist standard featur email messag word titl bodi sender receiv id induc social featur sender receiv messag enrich vector represent input svm classifi predict import level test messag obtain signific perform improv baselin system induc social featur experi multi user data collect obtain signific perform improv baselin system induc social featur experi multi user data collect rel error reduct mae micro averag macro averag 