Our research projects represent a comprehensive approach to decompositional semantics, each tackling a fundamental aspect of meaning representation through innovative annotation frameworks.
Decomp Toolkit
Comprehensive Python toolkit for working with Universal Decompositional Semantics datasets and models.
Explore ProjectDecompositional Parsing
Joint syntactic and semantic parsing models that learn to extract graph structures and attributes from text.
Explore ProjectPredPatt
Universal predicate-argument extraction that bridges deep syntax and shallow semantics across languages.
Explore ProjectSemantic Proto-Roles
Fine-grained thematic role properties that capture how events relate to their participants through decomposed semantic features.
Explore ProjectFactuality
Determining whether events described in text actually happened, incorporating speaker confidence and evidentiality.
Explore ProjectGenericity
Distinguishing generic statements about kinds from specific episodic events through multi-dimensional annotation.
Explore ProjectEvent Structure
Capturing complex event structures including subparts, participants, and their relationships over time.
Explore ProjectTime
Real-valued temporal relations and event durations for constructing precise document-level timelines.
Explore ProjectWord Sense
Word sense disambiguation through simple binary questions about meaning in context.
Explore Project