Abstract:In the real world, multimodal sentiment analysis (MSA) enables the capture and analysis of sentiments by fusing multimodal information, thereby enhancing the understanding of real-world environments. The key challenges lie in handling the noise in the acquired data and achieving effective multimodal fusion. When processing the noise in data, existing methods utilize the combination of multimodal features to mitigate errors in sentiment word recognition caused by the performance limitations of automatic speech recognition (ASR) models. However, there still remains the problem of how to more efficiently utilize and combine different modalities to address the data noise. In multimodal fusion, most existing fusion methods have limited adaptability to the feature differences between modalities, making it difficult to capture the potential complex nonlinear interactions that may exist between modalities. To overcome the aforementioned issues, this paper proposes a new framework named multimodal-word-refinement and cross-modal-hierarchy (MWRCMH) fusion. Specifically, we utilized a multimodal word correction module to reduce sentiment word recognition errors caused by ASR. During multimodal fusion, we designed a cross-modal hierarchical fusion module that employed cross-modal attention mechanisms to fuse features between pairs of modalities, resulting in fused bimodal-feature information. Then, the obtained bimodal information and the unimodal information were fused through the nonlinear layer to obtain the final multimodal sentiment feature information. Experimental results on the MOSI-SpeechBrain, MOSI-IBM, and MOSI-iFlytek datasets demonstrated that the proposed approach outperformed other comparative methods, achieving Has0-F1 scores of 76.43%, 80.15%, and 81.93%, respectively. Our approach exhibited better performance, as compared to multiple baselines.

Multimodal Sentiment Analysis Based on Cross-Modal Attention and Gated Cyclic Hierarchical Fusion Networks

Modality-invariant Temporal Representation Learning for Multimodal Sentiment Classification

Multimodal Sentiment Analysis Network Based on Distributional Transformation and Gated Cross-Modal Fusion

A cross modal hierarchical fusion multimodal sentiment analysis method based on multi-task learning

Multimodal Sentiment Analysis Based on Attentional Temporal Convolutional Network and Multi-Layer Feature Fusion

Multimodal Sentiment Analysis in Realistic Environments Based on Cross-Modal Hierarchical Fusion Network

Sentiment Analysis of Social Media Comments Based on Multimodal Attention Fusion Network

Multi-Channel Attentive Graph Convolutional Network with Sentiment Fusion for Multimodal Sentiment Analysis

Multimodal Sentiment Analysis Based on a Cross-Modal Multihead Attention Mechanism

A Graph Convolution-Based Heterogeneous Fusion Network for Multimodal Sentiment Analysis

Heterogeneous Hierarchical Fusion Network for Multimodal Sentiment Analysis in Real-World Environments

Context-Dependent Multimodal Sentiment Analysis Based on a Complex Attention Mechanism

Multi-Feature Fusion Multi-Modal Sentiment Analysis Model Based on Cross-Attention Mechanism

Multimodal Sentiment Analysis Using Multi-tensor Fusion Network with Cross-modal Modeling

Multimodal Sentiment Analysis Method Based on Hierarchical Adaptive Feature Fusion Network

Multimodal Sentiment Analysis with Word-Level Fusion and Reinforcement Learning

Multimodal Sentiment Analysis Based on Adaptive Gated Information Fusion

Multimodal Sentiment Analysis Based on Composite Hierarchical Fusion

Multimodal sentiment analysis based on multiple attention

Dynamically Shifting Multimodal Representations Via Hybrid-Modal Attention for Multimodal Sentiment Analysis

A transformer-encoder-based multimodal multi-attention fusion network for sentiment analysis