Thuật ngữ AI
Từ điển đầy đủ về Trí tuệ nhân tạo
Cross-modal alignment
Technique for matching semantic features between different modalities to ensure analysis consistency.
Multimodal latent representation
Compressed vector space where information from different modalities is fused into a unified semantic representation.
Multimodal transformer models
Transformer-based architectures capable of simultaneously processing multiple modalities through cross-attention mechanisms.
Sentiment dissonance
Detection of contradictions between sentiments expressed in different modalities (e.g., positive text with negative tone).
Modal weighting
Strategy for assigning different weights to each modality based on their relevance for sentiment prediction.
Multimodal contrastive learning
Training method that maximizes similarity between representations of the same sentiment and minimizes it between opposite sentiments.
Shared space projection
Technique for projecting features from different modalities into a common vector space to facilitate their comparison.
Multimodal anchor vector
Reference point in the latent space used to align representations from different modalities for sentiment.
Sentiment Co-occurrence
Statistical analysis of the simultaneous frequency of occurrence of specific sentiments across different modalities.
Trimodal Models
Architectures designed to simultaneously process three modalities (typically text, audio, and video) in sentiment analysis.