Nowadays,　in　the　big　data　era,　data　processing　technology　facilitates　us　to　get　the　utmost　out　of　sufficient　information　in　the　data.　However,　few　scholars　apply　big　data　technology　to　the　field　of　policy　evaluation.　Therefore,　under　the　Policy　Modeling　Consistency(PMC)　index　model　framework,　this　paper　thoroughly　mines　the　valuable　information　in　policy　texts　and　proposes　a　modular　policy　evaluation　system　combining　text　mining　and　machine　learning　methods.　The　system　is　divided　into　four　processing　modules,　including　data　acquisition,　data　processing,　index　evaluation　construction,　and　score　evaluation.　Compared　with　the　traditional　policy　evaluation　methods,　the　modular　policy　evaluation　system　presents　the　advantages　of　objectivity,　high　accuracy,　and　high　efficiency,　assisting　in　government　policies＇　implementation.