Rapid　recognition　of　depression　is　an　important　step　in　the　research　of　depression.　With　the　development　of　social　networking　platform,　more　and　more　depressive　patients　regard　micro-blog　as　one　of　the　ways　of　self-expression.　And　this　information　provides　support　of　data　for　the　recognition　of　depression.　In　this　study,　the　data　crawled　from　micro-blog＇s　＂tree　hole＂[1]　is　used　as　experimental　corpus.　Combined　with　the　features　of　micro-blog　text　with　depression,　a　double-input　convolutional　neural　network　structure　(D-CNN)　is　proposed.　This　method　takes　both　the　external　features　and　the　semantic　features　of　text　as　input.　By　comparing　the　accuracy　of　classification　with　Support　Vector　Machine　(SVM)　and　convolutional　neural　network　(CNN)　algorithm,　it　is　finally　shown　that　the　D-CNN　can　further　improve　the　accuracy　of　text　classify.