Semantic Representation And Composition Frameworks

via “compositional-visual-understanding-through-structured-annotations”

108K images with dense scene graphs and 5.4M region descriptions.

Unique: Provides explicit decomposition of images into objects, attributes, and relationships, enabling training of compositional models that understand visual scenes through structured components. Scene graphs naturally support compositional learning by representing images as compositions of objects and relationships.

vs others: Enables compositional learning unlike flat image-label datasets; supports training models that generalize to novel combinations of known components


Unique: Integrates formal semantic theory (first-order logic, lambda calculus) with computational approaches to meaning representation, showing how linguistic semantic phenomena map to computational structures. Includes discussion of semantic composition and how word meanings combine into sentence meanings.

vs others: More rigorous in formal semantic treatment than practical NLP guides, with deeper coverage of semantic phenomena (quantification, presupposition, negation) than most modern resources, making it essential for systems requiring semantic understanding beyond surface patterns.