Layers

Composable linguistic annotation lexicons for ATProto

Documentation • Discussions • Bluesky

What is Layers?

Layers is a set of AT Protocol Lexicon schemas under the pub.layers.* namespace. It defines a composable interchange format for linguistic annotations across text, audio, video, and image modalities, psycholinguistic and neurolinguistic signals, and offline judgment and experiment data.

Layers subsumes 15+ major annotation data models (CoNLL, CoNLL-U, brat, ELAN, TEI, WebVTT, Universal Dependencies, AMT, SRL, ARK, and others) while maintaining a theory-neutral, modular architecture. All data lives in user-controlled Personal Data Servers (PDSes); Layers provides the schema and protocols for interoperability.

Architecture overview

Layers is organized around a core pipeline of annotation layers, with parallel tracks for experimental and analytical workflows and integration layers connecting to the ATProto ecosystem:

Core Pipeline
  Expression (any linguistic unit: document, paragraph, sentence, word, morpheme)
         ↓
  Segmentation (tokenization strategies, token sequences)
         ↓
  Annotation (linguistic labels: POS, NER, semantic roles, etc.)

Parallel Tracks
  Ontology · Corpus · Resource · Judgment · Alignment

Integration Layers
  Graph · Eprint · Media · Persona · Changelog

See the documentation for the full architecture, including dependency graphs and cross-referencing patterns.

Lexicons

14 lexicon namespaces define 26 record types and 90 lexicon schemas:

Namespace	Purpose
pub.layers.defs	Shared primitives: anchors, selectors, metadata, cross-references
pub.layers.expression	Recursive document and expression model
pub.layers.segmentation	Tokenization strategies and token sequences
pub.layers.annotation	Annotation layers and cluster sets
pub.layers.ontology	Annotation type systems, role slots, and theoretical frameworks
pub.layers.corpus	Corpus collections with annotation design metadata
pub.layers.resource	Lexical entries, stimulus templates, fillings, and collections
pub.layers.judgment	Experiment definitions, judgment sets, and agreement reports
pub.layers.alignment	Cross-lingual and cross-modal structure correspondence
pub.layers.graph	Typed property graph nodes, edges, and edge sets
pub.layers.persona	Annotator personas and annotation frameworks
pub.layers.media	Audio, video, image, and signal metadata
pub.layers.eprint	Scholarly metadata and data provenance links
pub.layers.changelog	Structured change tracking with sub-record targeting

Project structure

layers/
├── lexicons/               # ATProto lexicon schemas (pub.layers.*)
│   └── pub/layers/         # 14 namespace directories, 90 JSON files
│       ├── defs.json        # Shared primitives
│       ├── expression/      # Record types, queries, and namespace defs
│       ├── annotation/
│       ├── ...
│       └── changelog/
├── docs/                   # Docusaurus documentation site (docs.layers.pub)
│   ├── docs/               # Markdown source files
│   ├── docusaurus.config.ts
│   └── sidebars.ts
├── CHANGELOG.md
└── validate-lexicons.mjs   # Lexicon validation script

Validation

npm install
npm test

Contributing

This project is in the design phase. Open issues or discussions to provide feedback on the lexicon design.

License

Licensed under CC-BY-SA-4.0.