郑州大学学报（理学版）2025，Vol.57Issue(5)：16-23,8.DOI:10.13705/j.issn.1671-6841.2024096
融合语言特征的多模态中文反讽识别模型
A Multimodal Chinese Sarcasm Detection Model Integrated with Linguistic Features
摘要
Abstract
In light of the disparity between graphic and text modes and insufficient attention to textual in-formation,a multimodal Chinese sarcasm detection model integrated with linguistic features was pro-posed.The Chi-square statistical method was used to extract words with sarcastic and non-sarcastic mean-ings,forming the linguistic feature system.TextCNN was utilized to extract linguistic features,enhancing the distinction between sarcastic and non-sarcastic characteristics.TextCNN and ResNet were employed to extract text and image features,and a cross-attention mechanism was introduced.Residual connections were used to fuse text and image features,to help preserve language characteristics.The effectiveness of the proposed model was verified by using an emergency multimodal dataset containing sarcastic com-ments.The results showed that the model outperformed the baseline model,and focusing on textual lin-guistic features helped improve the efficiency of problem-solving.关键词
语言特征/中文反讽识别/突发事件/多模态Key words
linguistic feature/Chinese sarcasm detection/emergency event/multimodality分类
信息技术与安全科学引用本文复制引用
胡文彬,陈龙,韩天乐,仲兆满,马常霞..融合语言特征的多模态中文反讽识别模型[J].郑州大学学报（理学版）,2025,57(5):16-23,8.基金项目
国家自然科学基金项目(72174079) （72174079）
江苏省"青蓝工程"优秀教学团队项目(2022-29) （2022-29）
