[1]欧阳梦妮,樊小超,帕力旦·吐尔逊.基于目标对齐和语义过滤的多模态情感分析[J].计算机技术与发展,2024,34(10):171-177.[doi:10.20165/j.cnki.ISSN1673-629X.2024.0209]
　OUYANG Meng-ni,FAN Xiao-chao,Palidan Turson.Multimodal Sentiment Analysis Based on Target Alignment and Semantic Filtering[J].,2024,34(10):171-177.[doi:10.20165/j.cnki.ISSN1673-629X.2024.0209]
点击复制

基于目标对齐和语义过滤的多模态情感分析()

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:
34
期数:
2024年10期
页码:
171-177
栏目:
人工智能
出版日期:
2024-10-10

文章信息/Info

Title:
Multimodal Sentiment Analysis Based on Target Alignment and Semantic Filtering
文章编号:
1673-629X(2024)10-0171-07
作者:
欧阳梦妮; 樊小超; 帕力旦·吐尔逊
新疆师范大学 计算机科学技术学院,新疆 乌鲁木齐 830054
Author(s):
OUYANG Meng-ni; FAN Xiao-chao; Palidan Turson
School of Computer Science and Technology,Xinjiang Normal University,Urumqi 830054,China
关键词:
方面级情感分析; 目标对齐; 语义过滤; 噪声; 多模态
Keywords:
aspect-based sentiment analysis; target alignment; semantic filtering; noise; multimodal
分类号:
TP391
DOI:
10.20165/j.cnki.ISSN1673-629X.2024.0209
摘要:
近年来许多研究工作利用注意力机制捕捉意见目标相应的视觉表征进行情感预测,但这种方法在细粒度意见目标对齐方面效果并不理想。 为此,提出一种基于目标对齐和语义过滤的多模态情感分析方法。 首先,引入目标识别方法 Deepface 获取图像的粗粒度意见目标,并使用映射方法,将粗粒度意见目标映射到细粒度意见目标,实现模态内的目标对齐。 其次,利用 Deepface 获取粗粒度意见目标的情绪词并将其和视觉表征融合,使模型更准确地理解和表示意见目标的情感倾向。 最后,引入图文匹配模型 CLIP 来评估图像与意见目标之间的语义关联性,从而过滤多余的视觉模态数据噪声。 实验表明,提出的意见目标对齐和语义过滤能更好地利用视觉模态信息,提高情感预测的准确性。
Abstract:
In recent years,many studies have utilized attention mechanisms to capture visual representations corresponding to opinion targets for sentiment prediction,but such methods are not ideal for fine-grained opinion target alignment. To address this,a multimodal sentiment analysis method based on target alignment and semantic filtering is proposed. First,the target recognition method Deepface is introduced to obtain coarse-grained opinion targets from images,and a mapping method is used to map these coarse-grained opinion targets to fine-grained opinion targets,achieving intra-modal target alignment. Second,emotion words associated with coarse-grained opinion targets obtained by Deepface are fused with visual representations,enabling the model to more accurately understand and represent the emotional tendencies of opinion targets. Finally, the text - image matching model CLIP is introduced to evaluate the semantic correlation between images and opinion targets,thereby filtering out redundant visual modal data noise. Experiments demonstrate that the proposed opinion target alignment and semantic filtering can better utilize visual modal information and improve the accuracy of sentiment prediction.

相似文献/References:

[1]翟社平,成大宝,张文晴,等.一种融合位置特征的方面级情感分析方法[J].计算机技术与发展,2023,33(05):167.[doi:10. 3969 / j. issn. 1673-629X. 2023. 05. 025]
　ZHAI She-ping,CHENG Da-bao,ZHANG Wen-qing,et al.An Aspect-level Sentiment Analysis Method with Location Features[J].,2023,33(10):167.[doi:10. 3969 / j. issn. 1673-629X. 2023. 05. 025]
[2]刘 洋,黎茂锋,黄 俊,等.面向方面级情感分析的双通道图卷积网络[J].计算机技术与发展,2024,34(03):49.[doi:10. 3969 / j. issn. 1673-629X. 2024. 03. 008]
　LIU Yang,LI Mao-feng,HUANG Jun,et al.Aspect-based Sentiment Analysis via Dual-channel Graph Convolutional Network[J].,2024,34(10):49.[doi:10. 3969 / j. issn. 1673-629X. 2024. 03. 008]
[3]孙晓晔,成彬*,王程.基于ChatGLM3-6B的方面级情感分析研究[J].计算机技术与发展,2025,(05):106.[doi:10.20165/j.cnki.ISSN1673-629X.2024.0391]
　SUN Xiao-ye,CHENG Bin*,WANG Cheng.Research on Aspect-based Sentiment Analysis Based on ChatGLM3-6B[J].,2025,(10):106.[doi:10.20165/j.cnki.ISSN1673-629X.2024.0391]
[4]任淑霞,刘旭琴.结合外部知识重构依赖矩阵的交互式图卷积网络[J].计算机技术与发展,2024,34(07):116.[doi:10.20165/j.cnki.ISSN1673-629X.2024.0125]
　REN Shu-xia,LIU Xu-qin.An Interactive Graph Convolutional Network of Reconstructing Dependence Matrix with External Knowledge[J].,2024,34(10):116.[doi:10.20165/j.cnki.ISSN1673-629X.2024.0125]

更新日期/Last Update: 2024-10-10