Social　networks　have　become　indispensable　in　people＇s　lives.　Despite　the　conveniences　brought　by　social　networks,　the　fake　news　on　those　online　platforms　also　induces　negative　impacts　and　losses　for　users.　With　the　development　of　deep　learning　technologies,　detecting　fake　news　in　a　data-driven　manner　has　attracted　great　attention.　In　this　paper,　we　give　a　brief　survey　that　discusses　the　recent　development　of　deep　learning　methods　in　fake　news　detection.　Compared　with　previous　surveys,　we　focus　on　the　different　data　structures　instead　of　the　models　they　used　to　process　those　data.　We　give　a　new　taxonomy　that　categorizes　current　models　into　the　following　three　parts:　models　that　formulate　fake　news　detection　as　text　classification,　models　that　formulate　fake　news　detection　as　graph　classification,　and　models　that　formulate　fake　news　detection　as　hybrid　classification.　The　advantages　and　drawbacks　of　those　methods　are　also　discussed.　©　2022　The　Author(s).