Aiming　at　the　problem　that　it　is　difficult　to　efficiently　identify　homologous　complaints　in　complaint　reporting,　this　paper　proposes　a　Siamese　Text　Convolutional　Neural　Network　(TextCNN)　method　to　detect　homologous　complaints　in　complaint　reporting.　TextCNN　is　good　at　capturing　local　features　in　text,　while　two　TextCNN　with　the　same　structure　can　capture　similar　features　in　two　complaint　reporting　samples　and　judge　whether　the　two　complaint　reporting　events　are　homology　events　by　comparing　these　features.　Meanwhile,　attention　mechanism　is　used　to　enhance　model　performance.　This　paper　compares　the　accuracy　of　multiple　neural　networks　in　complaint　reporting　data,　and　the　experimental　results　of　these　neural　network　models　show　that　the　Siamese　TextCNN　model　in　this　paper　has　better　effect　and　stronger　stability.　©　2021　IEEE