NAF version 4 development

NAF (NLP Annotation Format) is an XML format for annotating text documents. Different layers of annotation (tokenization, POS tagging, etc.) are represented as standalone annotations in a same document, while information about the models or linguistic processors used to create these annotations are recorded in the document header, together with metadata about the document. See example files in ./examples.

NAF was originally developed as part of the Newsreader project. Documentation for NAF until version 3.1 is hosted in newsreader/NAF.

This repository presents current NAF development.

Version overview

• The newsreader/NAF repository documents NAF version 3 as used for the Newsreader project, and up to version 3.1.
• Version 3.1 is the first step in further developing NAF for the Dutch FrameNet project, for which we are building a data-to-text annotation tool.
• Version 3.2 adds a text-unit element. This was developed for processing VOC missives for Clariah+/Text
• Version 3.3 extends annotations in linguistic-processor elements.
  • The current version is 3.3.1 (chunk elements have one and only one span sub-element)

The DTD of each version can be found in ./resources/dtd, version changes are documented in ./doc and Changelog.md.

The files under ./examples can be validated with the scripts ./scripts/validate_against_dtd.py (python 3.6+) and ./scripts/tests/. You will need to install lxml and pytest for that:

pip install -r requirements.txt

To validate all examples:

pytest scripts/tests

License

This project is licensed under the Apache 2.0 License - see the LICENSE.md file for details