Loading [a11y]/accessibility-menu.js
An Interactive Attention Mechanism Fusion Network for Aspect-Based Multimodal Sentiment Analysis | IEEE Conference Publication | IEEE Xplore

An Interactive Attention Mechanism Fusion Network for Aspect-Based Multimodal Sentiment Analysis


Abstract:

The goal of aspect-based multimodal sentiment analysis (ABMSA) is to classify the sentiment associated with aspect words in a given context. Most current ABMSA models foc...Show More

Abstract:

The goal of aspect-based multimodal sentiment analysis (ABMSA) is to classify the sentiment associated with aspect words in a given context. Most current ABMSA models focus only on general inter-modal information interactions without considering both intra-modal and inter-modal information interactions and ignoring image noise. To address these issues, this paper proposes an Interactive Attention Mechanism Fusion Network (IAMFN) model. The model first designs an image-text fusion module based on the attention, which applies the attention mechanism to a recurrent neural network to fuse text and images while filtering the noise in the images, and finally adds the fused information to the aspect information step by step to obtain the dynamic inter-modal representation. In addition, this paper proposes an aspect-text fusion module based on the attention, which learns the intra-modal contextual representation by calculating the weights of each aspect word in the context. Finally, this paper stitches the information obtained from the two modules and feeds it into the fully connected softmax layers to predict sentiment polarity. We have conducted extensive experiments on two benchmark datasets and the experimental results show that our model achieves state-of-the-art performance.
Date of Conference: 09-11 July 2023
Date Added to IEEE Xplore: 28 November 2023
ISBN Information:

ISSN Information:

Conference Location: Adelaide, Australia

Funding Agency:


References

References is not available for this document.