Skip to main content

Data Model Integration

Each page in this section maps a specific annotation framework's data model to Layers lexicons, identifying which pub.layers.* types, fields, and patterns correspond to the framework's constructs.

The goal is not to replicate each framework's API or file format, but to show that Layers' abstract primitives can faithfully represent the data each framework produces. If you can export data from one of these frameworks, you can represent it in Layers without information loss.

Frameworks

FrameworkOriginPrimary Focus
ConcreteHLTCOE, Johns HopkinsMulti-layer NLP pipeline output
beadFACTS.lab, Johns HopkinsTemplate-based judgment experiments
FOVEAparafovea projectPersona-driven annotation with ontologies
LAF/GrAFISO 24612Stand-off graph annotation
UIMA/CASApache/OASISType system-driven analysis
CoNLLCoNLL shared tasksColumn-based token annotation
TEIText Encoding InitiativeXML document encoding
ELAN/PraatMPI / University of AmsterdamTime-aligned multimedia annotation
FoLiARadboud UniversityXML linguistic annotation
NAFNewsReader projectNLP pipeline interchange
bratTsujii Laboratory, University of TokyoWeb-based text annotation
AMR / UCCA / DRS / EDSVariousSemantic graph formalisms
PAULA/Salt/ANNISHumboldt University / corpus-tools.orgMulti-layer corpus architecture
NIFAKSW, University of LeipzigLinked Data NLP interchange
W3C Web AnnotationW3C RecommendationWeb-based annotation
Decomp / UDSDecompositional Semantics InitiativeReal-valued semantic property graphs

Methodology

Each integration page follows a consistent structure. It identifies the framework's core concepts, then maps each concept to the corresponding Layers type or pattern.

Where a framework's concept has no direct Layers equivalent, the document explains how to represent it using Layers' composable primitives (typically featureMap for framework-specific metadata, or knowledgeRefs for external grounding).