One-step DCCA fusing BERT text with audio and video embeddings outperforms prior multi-modal methods for sentiment classification on two benchmarks and a new Debate Emotion dataset.
Multimodal sentiment analysis: Addressin g key issues and setting up the baselines
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.IR 1years
2019 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Multi-modal Sentiment Analysis using Deep Canonical Correlation Analysis
One-step DCCA fusing BERT text with audio and video embeddings outperforms prior multi-modal methods for sentiment classification on two benchmarks and a new Debate Emotion dataset.