A Multi-modal Sentiment Analysis Model Based on RoBERTa-AOBERT

The paper introduces a method of improving the AOBERT model with the main concept of substituting the original model's BERT component with the more efficient RoBERTa model. This modification is intended to boost the expressiveness and efficiency of the AOBERT in natural language processing tasks. Results of experiments show that upon substituting the use of RoBERTa in place of BERT, the model exhibits considerable improvement in various metrics. This study not only confirms the utility of the use of RoBERTa in the framework of AOBERT but also presents fresh insights and directions to be pursued in next-generation model improvement and expansion and may propel further innovation in related research areas.