Enhanced sentiment analysis in tourism reviews via multimodal graph convolutional networks

Abstract

In recent years, multimodal sentiment analysis has gained prominence due to its ability to leverage diverse data types for improved accuracy. However, combining text and image modalities presents challenges in effectively integrating and processing these diverse data sources for more effectively sentiment analysis. This study introduces a novel multimodal graph convolutional network GraphSAGE sentiment analysis model that integrates text and image features using advanced techniques include BERT for text graph construction and ResNet for image feature extraction and image graph construction, combined with a multi-head attention mechanism for effective edges fusion and nodes fusion. Experimental results on several datasets demonstrate that the proposed model outperforms existing state-of-the-art methods in terms of accuracy and efficiency. The findings highlight that the proposed model using multimodal data significantly enhance sentiment analysis performance. The source code for this study can be viewed at this url: https://github.com/CunyuanYang/Muilti-model_ESA .