Graph Fragment Language for Easy Syntactic Annotation
Python Other
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.

Graph Fragment Language for Easy Syntactic Annotation

Software to support a lightweight dependency-style annotation language. The notation is called GFL (Graph Fragment Language), and the formalism it expresses is called FUDG (Fragmentary Unlabeled Dependency Grammar). These are introduced in

A Framework for (Under)specifying Dependency Syntax without Overloading Annotators
Nathan Schneider, Brendan O’Connor, Naomi Saphra, David Bamman, Manaal Faruqui, Noah A. Smith, Chris Dyer, and Jason Baldridge. In Proceedings of the 7th Linguistic Annotation Workshop & Interoperability with Discourse, Sofia, Bulgaria, August 2013.

This repository contains supporting software developed by Nathan Schneider, Brendan O'Connor, Naomi Saphra, and Chris Dyer.

  • gflparser/ -- a parser for the GFL annotations format. It is implemented with a PEG (Parsing Expression Grammar).
    • parser/ contains code for a deprecated version of the annotation parser.
  • scripts/ -- a tool for checking GFL annotations (using the annotation parser), visualizing them in GraphViz, and converting them to other formats (including JSON).
  • guidelines/ -- annotation guidelines.
  • other directories contain some annotated sentences.
  • -- describes conventions for converting GFL/FUDG graphs into conventional labeled dependency trees.

For more information on GFL/FUDG, including a web annotation tool, see:

A FUDG dependency parser for tweets, and the data on which it was trained (Tweeboparser and Tweebank; Kong et al., EMNLP 2014) are available at:

Getting started

The viewer needs Python 2.7 (including version 0.3 of the parsimonious library—more recent versions are known to cause problems) and GraphViz (dot command) to be installed. Run:

scripts/ anno/tweets/dev.0000.anno
scripts/ -w anno/tweets/dev.0000.anno
scripts/ anno/nietzsche.anno