With　the　popularity　of　Internet　video,　video　retrieval　applications　become　more　and　more　widespread.　The　effect　of　the　traditional　retrieval　optimization　algorithms　cannot　meet　the　needs　of　users.　To　improve　rearrangement　semantic　rationality　of　video　search　results,　this　paper　introduces　the　video　annotation　to　mark　based　on　the　video　content　objectively.　At　the　same　time,　the　use　of　words　semantic　knowledge　tree　is　important.　Use　the　video　annotation　text　and　search　terms　in　order　to　quantify　the　performance　of　the　similarity.　And　combine　video　title　and　related　text　information　to　get　the　final　video　reranking　sequence.　In　this　paper,　the　validity　was　validated　by　experimental　method,　this　paper　also　add　video　test　on　network　video.　Experimental　results　show　that　the　performance　of　the　method　of　reranking　is　improved.