This repository contains code and annotations for the paper
Teresa Bürkle, Stefan Grünewald, and Annemarie Friedrich (2021): A Corpus Study of Creating Rule-Based Enhanced Universal Dependencies for German
The resources allow users to reproduce and extend the results reported in the study. Please cite the above paper when using them, and direct any questions or feedback regarding our system at Teresa Bürkle.
This software is a research prototype, solely developed for and published as part of the publication cited above. It will neither be maintained nor monitored in any way.
The converter.py contains the code for the converter. It takes a .conllu-file containing basic annotations as input
and produces an enhanced representation of those sentences by applying a set of hand-written rules.
Adds enhancements for the three phenomena coordination, relative clauses and raising/control.
basic_file- path to conllu-file containing basic annotations (required)enhanced_file- path to output file (required)--use_xsubj- option to use the relation subtypensubj:xsubjinstead of the generalnsubjrelation when enhancing raising / control constructions (default=False)
The annotations folder contains the manually annotated enhanced data we evaluated our converter against. The sentences for evaluation were taken from the treebanks GSD and PUD.
| File | Description |
|---|---|
| basic.conllu | basic layer for sentences used for evaluation, including basic layer corrections |
| gold.conllu | manually annotated/corrected enhanced layer for sentences used for evaluation |
| converted_enhanced.conllu | automatically converted version of basic.conllu using converter.py in code folder |
| engl_enhanced.conllu | automatically converted version of basic.conllu using slightly modified version of the english converter (Schuster & Manning 2016) |
| parsedbasic.conllu | basic layer for sentences used for evaluation that has been parsed automatically from raw text input |
| parsedbasic_enhanced.conllu | automatically converted verison of parsedbasic.conllu using converter.py in code folder |
| end-to-end-parser_enhanced.conllu | automatically parsed enhanced layer from raw text using a parser that has been trained on automatically enhanced version of GSD (created by using converter.py on GSD |
Our code is open-sourced under the BSD 3-Clause license. See the LICENSE_code file for details.
Our annotations are released under the CC-BY-SA 4.0 license. See the LICENSE_annotations file for details.
The software, including its dependencies, may be covered by third party rights, including patents. You should not execute this code unless you have obtained the appropriate rights, which the authors are not purporting to give.