Study of Multimodal Information Mining and Semantic Understanding Technology Based on Deep Learning

Le résumé fourni par la source

Based on the deep learning of multimodal information mining and semantic understanding technology, this paper designs a sentiment analysis and entity recognition model with text and image multimodal information for the task. It enhances the accuracy and semantic understanding capability of the task by fusing text and image features. These experimental results show that, compared to the unimodal model, the multimodal model improves the accuracy and F1-score in the sentiment analysis task by 5.8% and 6.9%, respectively; in the entity recognition task, it improves precision and recall rate by 7.7% and 10.9%, respectively. These results have proven that the benefit of multi-modal information fusion can bring a better performance to complex scenarios and effectively mine the multi-level semantic information of the model. Based on this, the paper provided considerable support for promoting multi-modal deep learning models in real-world applications, especially for those tasks requiring information association across different modalities.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.