Textualized and Feature-based Models for Compound Multimodal Emotion Recognition in the Wild

TL;DR

This paper compares text-based and feature-based models for compound multimodal emotion recognition in videos, highlighting the potential and limitations of textualization of non-verbal cues using large language models.

Contribution

It introduces a textualization approach for multimodal ER, leveraging LLMs to encode non-verbal cues in text, and evaluates its effectiveness against traditional feature-based models.

Findings

Textualization performs worse than feature-based models on wild datasets with sparse transcripts.

Rich transcripts improve the accuracy of text-based models.

Feature-based models generally outperform textualization in challenging wild scenarios.

Abstract

Systems for multimodal emotion recognition (ER) are commonly trained to extract features from different modalities (e.g., visual, audio, and textual) that are combined to predict individual basic emotions. However, compound emotions often occur in real-world scenarios, and the uncertainty of recognizing such complex emotions over diverse modalities is challenging for feature-based models. As an alternative, emerging large language models (LLMs) like BERT and LLaMA can rely on explicit non-verbal cues that may be translated from different non-textual modalities (e.g., audio and visual) into text. Textualization of modalities augments data with emotional cues to help the LLM encode the interconnections between all modalities in a shared text space. In such text-based models, prior knowledge of ER tasks is leveraged to textualize relevant non-verbal cues such as audio tone from vocal expressions, and action unit intensity from facial expressions. Since the pre-trained weights are publicly available for many LLMs, training on large-scale datasets is unnecessary, allowing to fine-tune for downstream tasks such as compound ER (CER). This paper compares the potential of text- and feature-based approaches for compound multimodal ER in videos. Experiments were conducted on the challenging C-EXPR-DB dataset in the wild for CER, and contrasted with results on the MELD dataset for basic ER. Our results indicate that multimodal textualization provides lower accuracy than feature-based models on C-EXPR-DB, where text transcripts are captured in the wild. However, higher accuracy can be achieved when the video data has rich transcripts. Our code is available.