Multimodal Sentiment Analysis: Addressing Key Issues and Setting Up the Baselines Setting up the baseline in 2018 on IEMOCAP 3DCNN and openSMILE for visual and acoustic feature extraction, then concatenate the feature vectors. Baseline method: BiLSTM to capture the context SVM with RBF kernel, thus no context Speaker-exclusive IEMOCAP: exclude one speaker at a time … Continue reading [GSoC2020]Audio-visual on IEMOCAP