Details

Autor(en) / Beteiligte

• Wang, Hejian
• Chen, Peng
• Guo, Qinglei
• Zhao, Lihua
• Ma, Xiaoxiao
• Li, Xuefeng
• Gao, Bo
• Li, Yongliang

Titel

False Information Detection Based on Multimodal Attention Detection Network

Ist Teil von

• 2023 6th International Conference on Artificial Intelligence and Big Data (ICAIBD), 2023, p.530-534

Ort / Verlag

IEEE

Erscheinungsjahr

2023

Quelle

IEEE Electronic Library (IEL)

Beschreibungen/Notizen

• With the development of the Internet, the distance between people is getting closer and closer under the bridge of the Internet. Especially the emergence of social software and media networks such as Weibo, Twitter, etc., has truly realized that "scholars do not go out and know everything about the World". However, this also provides a way for criminals to spread false information. False information mostly appears on the Internet in the form of a combination of pictures and texts, which makes it very difficult to detect and identify false information. The current mainstream multi-modal false information detection methods do not consider the relationship between different modal features, and more only use simple feature splicing to detect and discriminate false information, and the detection effect is very poor. Aiming at this problem, this paper proposes a multimodal false information detection method based on attention mechanism. After extracting the feature vectors of the visual and language modalities, the attention mechanism is used to realize the interaction between different modal features, and the model construction is completed through feature fusion. Finally, the detection of false information is realized. Compared with the current mainstream methods, the model proposed in this paper has achieved an excellent improvement, with an improvement of more than 6% in terms of accuracy.