notesum.ai

Published at December 10

Multimodal Sentiment Analysis Based on Causal Reasoning

cs.MM

cs.CL

Released Date: December 10, 2024

Authors: Fuhai Chen¹, Pengpeng Huang, Xuri Ge, Jie Huang, Zishuo Bao

Aff.: ¹University

Arxiv: http://arxiv.org/pdf/2412.07292v1

Refer to caption

Method	Condition	MVSA-Single		MVSA-Multiple
Method	Condition	Accuracy	F1	Accuracy	F1
OTE	Multimodal Biased	73.24	72.93	65.40	64.78
CMAC	Multimodal Biased	68.55	64.27	64.44	61.58
HSTEC	Multimodal Biased	71.60	67.49	63.25	60.27
NaiveCat	Multimodal Biased	72.46	71.79	65.52	65.65
NaiveCombine	Multimodal Biased	66.02	65.93	55.81	49.27
Baseline	Multimodal Biased	73.63	70.90	65.63	64.78
\hdashlineCF-MSA (Ours)	Removing Text Bias	76.17	74.54	67.12	64.92
	Removing Image Bias	73.24	72.32	66.23	64.53
	Removing Text-Image Bias	74.80	74.13	65.75	64.99