YZ Sözlüğü
Yapay Zekanın tam sözlüğü
Modality Fission
Inverse mechanism of fusion, consisting of decomposing an internal semantic representation into several output signals adapted to each modality (text generation, speech synthesis, image display).
Shared Latent Representation
Unique, reduced-dimensional vector space where features from different modalities are encoded, facilitating translation and comparison tasks between them.
Situated Dialogue
Form of dialogue where context is enriched and anchored in a physical or perceptible environment (e.g., a robot in a room), requiring continuous interpretation of sensory signals.
Multimodal Intent Management
Detection and interpretation of user intent by combining cues from multiple channels, such as a hand gesture accompanied by a voice command to designate an object.
Modality Correlation
Statistical analysis or learning of relationships and dependencies between data streams from different modalities to improve system coherence and accuracy.
Early Fusion Architecture
Integration strategy where raw features from each modality are concatenated or combined at a very early stage of the model, before deep processing layers.
Late Fusion Architecture
Approach where each modality is processed by a distinct sub-network until an advanced stage, with the outputs of these sub-networks then fused for a final decision.