Dynamic Tuning and Multi-Task Learning-Based Model for Multimodal Sentiment Analysis

Multimodal sentiment analysis aims to uncover human affective states by integrating data from multiple sensory sources. However, previous studies have focused on optimizing model architecture, neglecting the impact of objective function settings on model performance. Given this, this study introduce...

Full description

Saved in:
Bibliographic Details
Main Authors: Yi Liang, Turdi Tohti, Wenpeng Hu, Bo Kong, Dongfang Han, Tianwei Yan, Askar Hamdulla
Format: Article
Language:English
Published: MDPI AG 2025-06-01
Series:Applied Sciences
Subjects:
Online Access:https://www.mdpi.com/2076-3417/15/11/6342
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849330637470695424
author Yi Liang
Turdi Tohti
Wenpeng Hu
Bo Kong
Dongfang Han
Tianwei Yan
Askar Hamdulla
author_facet Yi Liang
Turdi Tohti
Wenpeng Hu
Bo Kong
Dongfang Han
Tianwei Yan
Askar Hamdulla
author_sort Yi Liang
collection DOAJ
description Multimodal sentiment analysis aims to uncover human affective states by integrating data from multiple sensory sources. However, previous studies have focused on optimizing model architecture, neglecting the impact of objective function settings on model performance. Given this, this study introduces a new framework, DMMSA, which utilizes the intrinsic correlation of sentiment signals and enhances the model’s understanding of complex sentiments. DMMSA incorporates coarse-grained sentiment analysis to reduce task complexity. Meanwhile, it embeds a contrastive learning mechanism within the modality, which decomposes unimodal features into similar and dissimilar ones, thus allowing for the simultaneous consideration of both unimodal and multimodal emotions. We tested DMMSA on the CH-SIMS, MOSI, and MOEI datasets. When only changing the optimization objectives, DMMSA achieved accuracy gains of 3.2%, 1.57%, and 1.95% over the baseline in five-class and seven-class classification tasks. In regression tasks, DMMSA reduced the Mean Absolute Error (MAE) by 1.46%, 1.5%, and 2.8% compared to the baseline.
format Article
id doaj-art-2d0ddb57b7904b539d2ede4e04445c0b
institution Kabale University
issn 2076-3417
language English
publishDate 2025-06-01
publisher MDPI AG
record_format Article
series Applied Sciences
spelling doaj-art-2d0ddb57b7904b539d2ede4e04445c0b2025-08-20T03:46:50ZengMDPI AGApplied Sciences2076-34172025-06-011511634210.3390/app15116342Dynamic Tuning and Multi-Task Learning-Based Model for Multimodal Sentiment AnalysisYi Liang0Turdi Tohti1Wenpeng Hu2Bo Kong3Dongfang Han4Tianwei Yan5Askar Hamdulla6School of Computer Science and Technology, Xinjiang University, Urumqi 830017, ChinaSchool of Computer Science and Technology, Xinjiang University, Urumqi 830017, ChinaInformation Research Center of Military Science, Beijing 100142, ChinaSchool of Computer Science and Technology, Xinjiang University, Urumqi 830017, ChinaSchool of Computer Science and Technology, Xinjiang University, Urumqi 830017, ChinaInformation Research Center of Military Science, Beijing 100142, ChinaSchool of Computer Science and Technology, Xinjiang University, Urumqi 830017, ChinaMultimodal sentiment analysis aims to uncover human affective states by integrating data from multiple sensory sources. However, previous studies have focused on optimizing model architecture, neglecting the impact of objective function settings on model performance. Given this, this study introduces a new framework, DMMSA, which utilizes the intrinsic correlation of sentiment signals and enhances the model’s understanding of complex sentiments. DMMSA incorporates coarse-grained sentiment analysis to reduce task complexity. Meanwhile, it embeds a contrastive learning mechanism within the modality, which decomposes unimodal features into similar and dissimilar ones, thus allowing for the simultaneous consideration of both unimodal and multimodal emotions. We tested DMMSA on the CH-SIMS, MOSI, and MOEI datasets. When only changing the optimization objectives, DMMSA achieved accuracy gains of 3.2%, 1.57%, and 1.95% over the baseline in five-class and seven-class classification tasks. In regression tasks, DMMSA reduced the Mean Absolute Error (MAE) by 1.46%, 1.5%, and 2.8% compared to the baseline.https://www.mdpi.com/2076-3417/15/11/6342multimodal sentiment analysiscontrastive learningmulti-task learningdynamic tuning
spellingShingle Yi Liang
Turdi Tohti
Wenpeng Hu
Bo Kong
Dongfang Han
Tianwei Yan
Askar Hamdulla
Dynamic Tuning and Multi-Task Learning-Based Model for Multimodal Sentiment Analysis
Applied Sciences
multimodal sentiment analysis
contrastive learning
multi-task learning
dynamic tuning
title Dynamic Tuning and Multi-Task Learning-Based Model for Multimodal Sentiment Analysis
title_full Dynamic Tuning and Multi-Task Learning-Based Model for Multimodal Sentiment Analysis
title_fullStr Dynamic Tuning and Multi-Task Learning-Based Model for Multimodal Sentiment Analysis
title_full_unstemmed Dynamic Tuning and Multi-Task Learning-Based Model for Multimodal Sentiment Analysis
title_short Dynamic Tuning and Multi-Task Learning-Based Model for Multimodal Sentiment Analysis
title_sort dynamic tuning and multi task learning based model for multimodal sentiment analysis
topic multimodal sentiment analysis
contrastive learning
multi-task learning
dynamic tuning
url https://www.mdpi.com/2076-3417/15/11/6342
work_keys_str_mv AT yiliang dynamictuningandmultitasklearningbasedmodelformultimodalsentimentanalysis
AT turditohti dynamictuningandmultitasklearningbasedmodelformultimodalsentimentanalysis
AT wenpenghu dynamictuningandmultitasklearningbasedmodelformultimodalsentimentanalysis
AT bokong dynamictuningandmultitasklearningbasedmodelformultimodalsentimentanalysis
AT dongfanghan dynamictuningandmultitasklearningbasedmodelformultimodalsentimentanalysis
AT tianweiyan dynamictuningandmultitasklearningbasedmodelformultimodalsentimentanalysis
AT askarhamdulla dynamictuningandmultitasklearningbasedmodelformultimodalsentimentanalysis