Decompositional
Semantics Initiative

Semantic annotation through simple, commonsensical questions.

Our Approach to Semantic Annotation

The Problem

Traditional semantic annotation frameworks require highly trained experts, complex category systems, and struggle with non-prototypical instances. They're expensive, slow, and often fail to capture the nuances of natural language.

Our Solution

Decompose complex semantic questions into simple, intuitive queries that anyone can answer. By breaking down meaning into its fundamental components, we achieve expert-level quality through crowd-sourcing while maintaining theoretical rigor.

Why Choose Decomp?

A semantic annotation paradigm that's simple and scalable

Simple Questions

Break down complex semantics into easy questions that anyone can answer. No linguistic expertise required.

Crowd-Scalable

Designed for massive annotation campaigns with built-in quality control and intuitive interfaces.

Theory-Driven

Grounded in decades of linguistic research while remaining accessible to everyone.

High Quality

Achieve annotation quality that rivals expert linguists through decomposition and aggregation.

Universal

Framework designed to work across languages and domains without modification.

Production Ready

Lightweight annotations perfect for real-world NLP systems and applications.

Featured Projects

View All Projects →

Get Started

Let's get decomposing!

Use the Toolkit

Get started with the Decomp toolkit - our comprehensive Python library for working with UDS data.

View on GitHub

Explore Data

Access our growing collection of high-quality semantic annotations across multiple languages and domains.

Browse Datasets

Collaborate

Join our community of researchers pushing the boundaries of semantic understanding.

Get in Touch