In　massive　text　information,　text　classification　can　better　help　people　organize　and　manage　the　mass　of　text　information.　In　real　life,　a　text　often　belongs　to　a　number　of　categories.　For　this　classification　scenario,　it　is　called　a　multi-label　classification.　Existing　text　multi-label　classification　methods　rarely　take　into　account　the　correlation　between　labels,　and　lack　the　understanding　of　the　label　semantics.　In　this　paper　we　propose　a　method　of　text　multi-label　learning　based　on　label　correlation,　modeling　between　texts　and　label　vector　through　neural　networks,　and　capture　the　semantic　correlation　between　text　and　labels.　A　large　number　of　experiments　show　the　effectiveness　of　the　method,　especially　when　the　amount　of　data　is　large.