Research and Application of Multimodal Feature Fusion in Short Text Classification using Dynamic Weighting Strategy

This paper explores the theory and practice of multimodal feature fusion in short text classification using dynamic weighting strategies. Aiming at the unstructured characteristics and information incompleteness of short text, this paper proposes an innovative classification method combined with deep learning technology, aiming to achieve more accurate and efficient classification. By constructing an advanced multimodal feature fusion model, the features of different modalities are integrated, and then richer semantic connotations are mined. Meanwhile, a dynamic weighting strategy is adopted to adaptively adjust the features of different modalities, ensuring that the model can adapt more flexibly to various data distributions. The experimental results verify the effectiveness and superiority of the method in short text categorisation, providing a solid theoretical and practical foundation for research in related fields.