Toprovidemoreaccurate,diverse,andexplainablerecommendation,itiscompulsorytogobeyondmodelinguser-iteminteractionsandtakesideinformationintoaccount.Traditionalmethodslikefactorizationmachine(FM)castitasasupervisedlearningproblem,whichassumeseachinteractionasanindependentinstancewithsidei