Table 1 High-level overview of modern approaches in multi-label sentiment classification.
From: A weighted difference loss approach for enhancing multi-label classification
Approach Category | Representative Methods / Key Papers | Core Strategy Keywords for Imbalance / Dependencies |
|---|---|---|
Loss Function Modification | Focal Loss25, Asymmetric Loss (ASL)26, Label Distribution Learning (LDL)27, Label Correlation Losses28 | Re-weighting, Decoupling, Distribution Learning, Direct Correlation Terms. |
Explicit Dependency Modeling | CRFs29, Label Embeddings (LEAM30), Graph Neural Networks (GNNs) (Structure-based31, KG-enhanced32) | Graphical Models, Semantic Embeddings, Graph Propagation, Knowledge Infusion. |
Advanced Representation | Contrastive Learning (Label-aware33), Prompt-based Learning (PLM/LLM Prompts34) | Discriminative Features, Text-Label Alignment, Task Reformulation, In-context Learning. |
