Multimodal Sarcasm Target Identification in Tweets

• DOI: 10.18653/v1/2022.acl-long.562
• Publication Date: 2022-06-03T01:34:53Z
• Authors: Jiquan Wang; Lin Sun; Yi Liu; Meizhi Shao; Zengwei Zheng
• Keywords: Sarcasm; Identification

Sarcasm is important to sentiment analysis on social media. Target Identification (STI) deserves further study understand sarcasm in depth. However, text lacking context or missing target makes identification very difficult. In this paper, we introduce multimodality STI and present Multimodal (MSTI) task. We propose a novel multi-scale cross-modality model that can simultaneously perform textual labeling visual detection. the model, extract features enrich spatial information for different sized targets. design set of convolution networks unify with cross-modal attention learning, correspondingly transposed restore information. The results show clues improve performance TSTI by large margin, VSTI achieves good accuracy.