turkish-amr-parser

This repository provides the resources introduced within the article:

Abstract Meaning Representation of Turkish accepted for publication in Natural Language Engineering, Cambridge Press.

Abstract meaning representation (AMR) is a graph-based sentence-level meaning representation. Although it is originally designed for English, its adaptation to non-English languages is possible by defining language-specific divergences and representations. We introduce the first AMR representation framework for Turkish; agglutinative, free constituent order, morphologically highly rich. Besides this main contribution, the article also presents the construction of the first AMR corpus of 700 sentences, the first AMR parser (i.e., a tree-to-graph rule-based AMR parser)

Other resources introduced within the same article may be found at the following GitHub repositories:

the first Turkish AMR Guidelines: turkish-amr-guidelines

the first Turkish AMR Corpus: turkish-amr-corpus

Quick Tour

Input Format

Turkish AMR parser uses syntactic and semantic features, therefore they should be provided to the parser in CONLL format.

A sample should have the following columns :

• Token id
• Surface form
• Lemma
• Pos
• PPos
• Morphologic Tags
• Dependency Head
• Dependency Relation
• SRL features

An example data sample: Ama annemin şartları vardı. (But my mother has her conditions)

ID	Surface	Lemma*	Pos	PPos	Morp*	DepHead ID	DepRel*	Semantic Layers
1	ama	ama	Conj	Conj	_	4	CONJUNCTION	_ _ _
2	annemin	anne	Noun	Noun	A3sg|P1sg|Gen	3	POSSESSOR	_ _ _
3	şartları	şart	Noun	Noun	A3pl|P3sg|Nom	4	SUBJECT	_ _ A1
4	vardı	var	Verb	Verb	Pos|Past|A3sg	0	PREDICATE	Y var.01 _
5	.	.	Punc	Punc	_	4	PUNCTUATION	_ _ _

* indicates duplicate columns. The further information about input form is available in Abstract Meaning Representation of Turkish

Running the Parser

For Prepocessing python preprocess.py -f dataset.conll --out ./output

Output file is save as propbank.out in output directory

For Parsing python parsing.py -f ./output/propbank.out --out ./parser_out

Parsing results are saved in ./parser_out/output.out file. Exceptions are also save in ./parser_out/exceptions.out.

To parse only one sentence

python parsing.py -f ./output/propbank.out --out ./parser_out --only-one

Contributing

All contributions welcome!

How to cite this work

Oral, E., Acar, A., & Eryiğit, G. (2022). Abstract Meaning Representation of Turkish. Natural Language Engineering, 1-33. doi:coming

Licence Information

Shield:

This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.