VLP2MSA: Expanding vision-language pre-training to multimodal sentiment analysis

Multimodal Aspect-Based Sentiment Analysis (MABSA) is an essential task in sentiment analysis that has garnered considerable attention in recent years. Typical approaches in MABSA often utilize cross-modal Transformers to capture interactions between textual and visual modalities. However, bridging the semantic gap between modalities spaces and addressing interference from irrelevant visual objects at different scales remains challenging. To tackle these limitations, we present the Multi-level Textual-Visual Alignment and Fusion Network (MTVAF) in this work, which incorporates three auxiliary tasks. Specifically, MTVAF first transforms multi-level image information into image descriptions, facial descriptions, and optical characters. These are then concatenated with the textual input to form a textual+visual input, facilitating comprehensive alignment between visual and textual modalities. Next, both inputs are fed into an integrated text model that incorporates relevant visual representations. Dynamic attention mechanisms are employed to generate visual prompts to control cross-modal fusion. Finally, we align the probability distributions of the textual input space and the textual+visual input space, effectively reducing noise introduced during the alignment process. Experimental results on two MABSA benchmark datasets demonstrate the effectiveness of the proposed MTVAF, showcasing its superior performance compared to state-of-the-art approaches. Our codes are available at https://github.com/MKMaS-GUET/MTVAF.

Multimodal sentiment analysis is an extended approach to traditional language-based sentiment analysis, which uses other relevant modality data. Multimodal sentiment analysis usually applies visual, textual, and acoustic representations for sentiment prediction. Recently, various data fusion methodologies have been proposed for multimodal sentiment analysis. In most cases, textual modality plays a major role, and visual and acoustic modalities are used as auxiliary sources for multimodal sentiment analysis. However, in general multimedia such as video, text transcripts of an individual's speech are not provided. Research on an audio-visual sentiment analysis methodology that does not depend on text modality is essential for multimodal sentiment analysis in real-world industrial applications. Therefore, it is important to improve audio-visual sentiment analysis because it currently exhibits lower performance than multimodal sentiment analysis, including text modality. In this paper, we propose heterogeneous modality transfer learning (HMTL) to utilize the knowledge of aligned text data as a source modality in transfer learning to improve audio-visual sentiment analysis performance. Our approach uses a decoder and adversarial learning techniques to reduce the gap between the source and target modalities in the embedded space for multimodal representation. Our proposed methodology experimentally outperformed recent unimodal and bimodal audio-visual sentiment analysis achievements.

TCMT: Target-oriented Cross Modal Transformer for Multimodal Aspect-Based Sentiment Analysis

With the increasing popularity of video sharing websites such as YouTube and Facebook, multimodal sentiment analysis has received increasing attention from the scientific community. Contrary to previous works in multimodal sentiment analysis which focus on holistic information in speech segments such as bag of words representations and average facial expression intensity, we propose a novel deep architecture for multimodal sentiment analysis that is able to perform modality fusion at the word level. In this paper, we propose the Gated Multimodal Embedding LSTM with Temporal Attention (GME-LSTM(A)) model that is composed of 2 modules. The Gated Multimodal Embedding allows us to alleviate the difficulties of fusion when there are noisy modalities. The LSTM with Temporal Attention can perform word level fusion at a finer fusion resolution between the input modalities and attends to the most important time steps. As a result, the GME-LSTM(A) is able to better model the multimodal structure of speech through time and perform better sentiment comprehension. We demonstrate the effectiveness of this approach on the publicly-available Multimodal Corpus of Sentiment Intensity and Subjectivity Analysis (CMU-MOSI) dataset by achieving state-of-the-art sentiment classification and regression results. Qualitative analysis on our model emphasizes the importance of the Temporal Attention Layer in sentiment prediction because the additional acoustic and visual modalities are noisy. We also demonstrate the effectiveness of the Gated Multimodal Embedding in selectively filtering these noisy modalities out. These results and analysis open new areas in the study of sentiment analysis in human communication and provide new models for multimodal fusion.

Integrating color cues to improve multimodal sentiment analysis in social media

Multimodal sentiment analysis is an important research area that predicts speaker’s sentiment tendency through features extracted from textual, visual and acoustic modalities. The central challenge is the fusion method of the multimodal information. A variety of fusion methods have been proposed, but few of them adopt end-to-end translation models to mine the subtle correlation between modalities. Enlightened by recent success of Transformer in the area of machine translation, we propose a new fusion method, TransModality, to address the task of multimodal sentiment analysis. We assume that translation between modalities contributes to a better joint representation of speaker’s utterance. With Transformer, the learned features embody the information both from the source modality and the target modality. We validate our model on multiple multimodal datasets: CMU-MOSI, MELD, IEMOCAP. The experiments show that our proposed method achieves the state-of-the-art performance.

Multimodal sentiment analysis (MSA) aims to use a variety of sensors to obtain and process information to predict the intensity and polarity of human emotions. The main challenges faced by current multi-modal sentiment analysis include: how the model extracts emotional information in a single modality and realizes the complementary transmission of multimodal information; how to output relatively stable predictions even when the sentiment embodied in a single modality is inconsistent with the multi-modal label; how can the model ensure high accuracy when a single modal information is incomplete or the feature extraction performance not good. Traditional methods do not take into account the interaction of unimodal contextual information and multi-modal information. They also ignore the independence and correlation of different modalities, which perform poorly when multimodal sentiment representations are asymmetric. To address these issues, this paper first proposes unimodal feature extraction network (UFEN) to extract unimodal features with stronger representation capabilities; then introduces multi-task fusion network (MTFN) to improve the correlation and fusion effect between multiple modalities. Multilayer feature extraction, attention mechanisms and Transformer are used in the model to mine potential relationships between features. Experimental results on MOSI, MOSEI and SIMS datasets show that the proposed method achieves better performance on multimodal sentiment analysis tasks compared with state-of-the-art baselines.

Sentiment-aware multimodal pre-training for multimodal sentiment analysis

The homogeneity and heterogeneity across modalities are critical factors that influence multimodal fusion. In Multimodal Sentiment Analysis (MSA), the inherent textual information within the audio modality induces cross-modality homogeneity with the text modality. Conversely, the mutual independence between text and vision modalities results in their cross-modal heterogeneity. Although existing disentangle-based methods achieve notable performance gains by separating modality features into distinct subspaces, they overlook the characteristics of cross-modality heterogeneity and homogeneity among different modalities. To this end, we propose a novel Modality-aware Disentangle and Fusion (MDF) framework to investigate the role of core modality features. Specifically, we first use text as the anchor to disentangle the audio modality and extract its unique modality-specific features, thereby establishing cross-modal heterogeneity among text, audio, and vision. We then introduce a Cross-Modality Heterogeneity Enhancement (CHE) module to refine these features, further reinforcing their heterogeneous characteristics. Finally, a Modality Adaptive Weighting (MAW) module is employed to dynamically assign weights to the text, sound, and vision modalities based on their potential contributions to sentiment prediction, achieving a more effective multimodal representation for MSA. Experimental evaluations on different benchmarks demonstrate MDF's superiority, with extensive ablation studies confirming its effectiveness.

Multimodal sentiment analysis has attracted increasing attention and lots of models have been proposed. However, the performance of the state-of-the-art models decreases sharply when they are deployed in the real world. We find that the main reason is that real-world applications can only access the text outputs by the automatic speech recognition (ASR) models, which may be with errors because of the limitation of model capacity. Through further analysis of the ASR outputs, we find that in some cases the sentiment words, the key sentiment elements in the textual modality, are recognized as other words, which makes the sentiment of the text change and hurts the performance of multimodal sentiment analysis models directly. To address this problem, we propose the sentiment word aware multimodal refinement model (SWRM), which can dynamically refine the erroneous sentiment words by leveraging multimodal sentiment clues. Specifically, we first use the sentiment word position detection module to obtain the most possible position of the sentiment word in the text and then utilize the multimodal sentiment word refinement module to dynamically refine the sentiment word embeddings. The refined embeddings are taken as the textual inputs of the multimodal feature fusion module to predict the sentiment labels. We conduct extensive experiments on the real-world datasets including MOSI-Speechbrain, MOSI-IBM, and MOSI-iFlytek and the results demonstrate the effectiveness of our model, which surpasses the current state-of-the-art models on three datasets. Furthermore, our approach can be adapted for other multimodal feature fusion models easily 1 .

Multimodal Sentiment Analysis (MSA) aims to predict subjective human emotions by leveraging multimodal information. However, existing research inadequately utilizes explicit sentiment semantic information at the lexical level in text and overlooks noise interference from non-dominant modalities, such as irrelevant movements in visual modalities and background noise in audio modalities. To address this issue, we propose a multimodal sentiment analysis model based on knowledge enhancement and text-guided learning (MMKT). The model constructs a sentiment knowledge graph for the textual modality using the SenticNet knowledge base. This graph directly annotates word-level sentiment polarity, strengthening the model’s understanding of emotional vocabulary. Furthermore, global sentiment knowledge features are generated through graph embedding computations to enhance the multimodal fusion process. Simultaneously, a dynamic text-guided learning approach is introduced, which dynamically leverages multi-scale textual features to actively suppress redundant or conflicting information in visual and audio modalities, thereby generating purer cross-modal representations. Finally, concatenated textual features, cross-modal features, and knowledge features are utilized for sentiment prediction. Experimental results on the CMU-MOSEI and Twitter2019 dataset demonstrate the superior performance of the MMKT model.

Multimodal sentiment analysis faces a number of challenges, including modality missing, modality heterogeneity gap, incomplete datasets, etc. Previous studies usually adopt schemes like meta-learning or multi-layer structures. Nevertheless, these methods lack interpretability for the interaction between modalities. In this paper, we constructed a new dataset, SM-MSD, for sentiment analysis in social media (SAS) that differs significantly from conventional corpora, comprising 10K instances of diverse data from Twitter, encompassing text, emoticons, emojis, and text embedded in images. This dataset aims to reflect authentic social scenarios and various emotional expressions, and provides a meaningful and challenging evaluation benchmark for multimodal sentiment analysis in specific contexts. Furthermore, we propose a multi-task framework based on heterogeneous graph neural networks (H-GNNs) and contrastive learning. For the first time, heterogeneous graph neural networks are applied to multimodal sentiment analysis tasks. In the case of additional labeling data, it guides the emotion prediction of the missing mode. We conduct extensive experiments on multiple datasets to verify the effectiveness of the proposed scheme. Experimental results demonstrate that our proposed scheme surpasses state-of-the-art methods by 1.7% and 0 in accuracy and 1.54% and 4.9% in F1-score on the MOSI and MOSEI datasets, respectively, and exhibits robustness to modality missing scenarios.

The wide application of smart devices enables the availability of multimodal data, which can be utilized in many tasks. In the field of multimodal sentiment analysis, most previous works focus on exploring intra- and inter-modal interactions. However, training a network with cross-modal information (language, audio and visual) is still challenging due to the modality gap. Besides, while learning dynamics within each sample draws great attention, the learning of inter-sample and inter-class relationships is neglected. Moreover, the size of datasets limits the generalization ability of the models. To address the afore-mentioned issues, we propose a novel framework HyCon for hybrid contrastive learning of tri-modal representation. Specifically, we simultaneously perform intra-/inter-modal contrastive learning and semi-contrastive learning, with which the model can fully explore cross-modal interactions, learn inter-sample and inter-class relationships, and reduce the modality gap. Besides, refinement term and modality margin are introduced to enable a better learning of unimodal pairs. Moreover, we devise pair selection mechanism to identify and assign weights to the informative negative and positive pairs. HyCon can naturally generate many training pairs for better generalization and reduce the negative effect of limited datasets. Extensive experiments demonstrate that our method outperforms baselines on multimodal sentiment analysis and emotion recognition.

Aiming at the existing multimodal sentiment analysis approaches, which include inadequate extraction of unimodal features, redundancy of independent modal features, insufficient analysis of semantic correlation between data and insufficient fusion, a Web-Semantic Enhanced Multimodal Sentiment Analysis Using Multilayer Cross-Attention Fusion is proposed. The model utilizes deep learning (including XLNet, ResNeSt, and convolutional neural networks) to extract high-level features from text, audio, and visual modes through self-attention mechanisms, and improves the accuracy of emotion classification through multimodal fusion. The results of experiments demonstrate that the suggested MCFMSA can achieve Acc-2, Acc-3, F1, and MAE values of 89.7%, 85.2%, 89.3%, and 0.466 on the CMU-MOSI dataset, respectively; and on the CMU-MOSEI dataset, Acc-2, Acc-3, F1, and MAE values of 88.7%, 82.5%, 86.5%, and 0.475. All of them are significantly improved compared to several other advanced multimodal sentiment analysis methods, which can enhance the accuracy of sentiment classification.

Multimodal sentiment analysis refers to the use of computers to analyze and identify the emotions that people want to express through the extracted multimodal sentiment features, and it plays a significant role in human-computer interaction and financial market prediction. Most existing approaches to multimodal sentiment analysis use contextual information for modeling, and while this modeling approach can effectively capture the contextual connections within modalities, the correlations between modalities are often overlooked, and the correlations between modalities are also critical to the final recognition results in multimodal sentiment analysis. Therefore, this paper proposes a multimodal sentiment analysis approach based on the universal transformer, a framework that uses the universal transformer to model the connections between multiple modalities while employing effective feature extraction methods to capture the contextual connections of individual modalities. We evaluated our proposed method on two benchmark datasets for multimodal sentiment analysis, CMU-MOSI and CMU-MOSEI, and the results outperformed other methods of the same type.