DWDSmor

SFST/SMOR/DWDS-based German morphology

This project provides a component for the lemmatisation and morphological analysis of word forms as well as for the generation of paradigms of lexical words in written German. To this end we adopt:

1. SFST, a C++ library and toolbox for finite-state transducers (FSTs) (Schmidt 2006)
2. SMORLemma (Sennrich and Kunz 2014), a modified version of the Stuttgart Morphology (SMOR) (Schmid, Fitschen, and Heid 2004) with an alternative lemmatisation component
3. the DWDS dictionary (BBAW n.d.) replacing the IMSLex (Fitschen 2004) as the lexical data source for German words, their grammatical categories, and their morphological properties.

This repository provides source code for building DWDSmor lexica and transducers as well as for using DWDSmor transducers for morphological analysis and paradigm generation:

• share/ contains XSLT stylesheets for extracting lexical entries in SMORLemma format form XML sources of DWDS articles. Sample inputs and outputs can be found in samples/.
• lexicon/dwds/ contains scripts for building DWDSmor lexica by means of the XSLT stylesheets in share/ and DWDS sources in lexicon/dwds/wb/, which are not part of this repository.
• lexicon/sample/ contains scripts for building sample DWDSmor lexica by means of the XSLT stylesheets in share/ and the sample lexicon in lexicon/sample/wb/.
• grammar/ contains an FST grammar derived from SMORLemma, providing the morphology for building DWDSmor automata from DWDSmor lexica.
• test/ implements a test suite for the DWDSmor transducers.
• dwdsmor.py and paradigm.py are user-level Python scripts for morphological analysis and for paradigm generation by means of DWDSmor transducers.

DWDSmor is in active development. In its current stage, DWDSmor supports most inflection classes and some productive word-formation patterns of written German. Note that the sample lexicon in lexicon/sample/wb/ only covers a sketchy subset of the German vocabulary, and so do the DWDSmor automata compiled from it.

Prerequisites

GNU/Linux : Development, builds and tests of DWDSmor are performed on Debian GNU/Linux. While other UNIX-like operating systems such as MacOS should work, too, they are not actively supported.

Python >= v3.9 : DWDSmor targets Python as its primary runtime environment. The DWDSmor transducers can be used via SFST's commandline tools, queried in Python applications via language-specific bindings, or used by the Python scripts dwdsmor.py and paradigm.py for morphological analysis and for paradigm generation.

Saxon-HE : The extraction of lexical entries from XML sources of DWDS articles is implemented in XSLT 2, for which Saxon-HE is used as the runtime environment.

Java (JDK) >= v8 : Saxon requires a Java runtime.

SFST : a C++ library and toolbox for finite-state transducers (FSTs); please take a look at its homepage for installation and usage instructions.

On a Debian-based distribution, install the following packages:

apt install python3 default-jdk libsaxonhe-java sfst

Set up a virtual environment for project builds via Python's venv:

python3 -m venv .venv
source .venv/bin/activate

Then run the DWDSmor setup routine in order to install Python dependencies:

make setup

Building DWDSmor lexica and transducers

For building DWDSmor lexica and transducers, run:

make all

Alternatively, you can run:

make dwds && make dwds-install && make dwdsmor

Note that these commands require DWDS sources in lexicon/dwds/wb/, which are not part of this repository.

Alternatively, you can build sample DWDSmor lexica and transducers from the sample lexicon in lexicon/sample/wb/ by running:

make sample && make sample-install && make dwdsmor

After building DWDSmor transducers, install them into lib/, where the user-level Python scripts dwdsmor.py and paradigm.py expect them by default:

make install

The installed DWDSmor transducers are:

• lib/dwdsmor.{a,ca}: transducer with inflection and word-formation components, for lemmatisation and morphological analysis of word forms in terms of grammatical categories
• lib/dwdsmor-finite.{a,ca}: transducer with an inflection component and a finite word-formation component, for testing purposes
• lib/dwdsmor-root.{a,ca}: transducer with inflection and word-formation components, for lexical analysis of word forms in terms of root lemmas (i.e., lemmas of ultimate word-formation bases), word-formation process, word-formation means, and grammatical categories in term of the Pattern-and-Restriction Theory of word formation (Nolda 2022)
• lib/dwdsmor-index.{a,ca}: transducer with an inflection component only with DWDS homographic lemma indices, for paradigm generation

Testing DWDSmor transducers

The installed DWDSmor transducers can be examined with the test suite in test/. It provides coverage tests and regression tests.

The coverage tests are run with the following command:

make test-coverage

Coverage test reports and statistics are saved as TSV tables in test/reports/ and test/summaries/, respectively.

Individual coverage tests can be run by calling test/Makefile as below:

make -C test test-dwds-lemma-coverage

make -C test test-sample-lemma-coverage

make -C test test-tuebadz-lemma-coverage

make -C test test-dwds-paradigm-coverage

make -C test test-sample-paradigm-coverage

The test-dwds-lemma-coverage and test-dwds-paradigm-coverage targets of test/Makefile require DWDS sources in lexicon/dwds/wb/ (not part of this repository). The test-tuebadz-lemma-coverage target presupposes a TüBa-D/Z treebank export tuebadz-11.0-exportXML-v2.xml at test/data/tuebadz/ (likewise not part of this repository).

Note that runs of the test-dwds-paradigm-coverage and test-sample-paradigm-coverage targets of test/Makefile may take a considerable amount of time.

Regression tests compare generated test results to saved snapshots in test/reports/. To create the snapshots, first run:

make test-snapshot

Then, in order to test for regressions which may arise from changes of lexicon, grammar, or user-level scripts, run:

make test-regression

Regression test targets can also be run individually by calling test/Makefile as follows:

make -C test test-analysis-snapshot

make -C test test-paradigm-snapshot

make -C test test-analysis-regression

make -C test test-paradigm-regression

Using DWDSmor

DWDSmor provides two Python scripts for using the DWDSmor transducers.

dwdsmor.py is a Python script for the lemmatisation and morphological analysis of word forms in written German by means of a DWDSmor transducer:

$ ./dwdsmor.py -h
usage: dwdsmor.py [-h] [-c] [-C] [-E] [-H] [-I] [-j] [-m] [-n] [-N] [-P] [-t TRANSDUCER]
                  [-v] [-W] [-y] [input] [output]

positional arguments:
  input                 input file (one word form per line; default: stdin)
  output                output file (default: stdout)

optional arguments:
  -h, --help            show this help message and exit
  -c, --csv             output CSV table
  -C, --force-color     preserve color and formatting when piping output
  -E, --no-empty        do not output empty columns or values
  -H, --no-header       suppress table header
  -I, --no-index        do not output lemma and paradigm index
  -j, --json            output JSON object
  -m, --minimal         prefer analyses with minimal number of stem boundaries
  -n, --no-analysis     do not output raw analysis
  -N, --no-segmentation
                        do not output segmented lemma
  -P, --plain           suppress color and formatting
  -t TRANSDUCER, --transducer TRANSDUCER
                        path to transducer file (default: lib/dwdsmor.ca)
  -v, --version         show program's version number and exit
  -W, --no-wf           do not output word-formation process and means
  -y, --yaml            output YAML document

By default, dwdsmor.py prints a TSV table on standard output:

$ echo "Ihr\nkönnt\neuch\nauf\nden\nKinderbänken\nausruhen\n." | ./dwdsmor.py -E -n
Wordform	Lemma	Segmentation	POS	Subcategory	Person	Gender	Case	Number	Inflection	Function	Nonfinite	Mood	Tense	Metalinguistic	Characters
Ihr	Sie	Sie	PPRO	Pers	3	NoGend	Gen	Pl						Old
Ihr	sie	sie	PPRO	Pers	3	NoGend	Gen	Pl						Old	CAP
Ihr	sie	sie	PPRO	Pers	3	Fem	Dat	Sg							CAP
Ihr	sie	sie	PPRO	Pers	3	Fem	Gen	Sg						Old	CAP
Ihr	ihr	ihr	PPRO	Pers	2		Nom	Pl							CAP
Ihr	ihre	ihre	POSS			Neut	Acc	Sg	NoInfl	Attr					CAP
Ihr	ihre	ihre	POSS			Neut	Nom	Sg	NoInfl	Attr					CAP
Ihr	ihre	ihre	POSS			Masc	Nom	Sg	NoInfl	Attr					CAP
Ihr	Ihre	Ihre	POSS			Neut	Acc	Sg	NoInfl	Attr
Ihr	Ihre	Ihre	POSS			Neut	Nom	Sg	NoInfl	Attr
Ihr	Ihre	Ihre	POSS			Masc	Nom	Sg	NoInfl	Attr
könnt	können	könn<~>en	V		2			Pl				Ind	Pres
euch	euch	euch	PPRO	Refl	2		Acc	Pl
euch	euch	euch	PPRO	Refl	2		Dat	Pl
euch	ihr	ihr	PPRO	Pers	2		Acc	Pl
euch	ihr	ihr	PPRO	Pers	2		Dat	Pl
auf	auf	auf	ADV
auf	auf	auf	PREP
den	die	die	REL			Masc	Acc	Sg	St	Subst
den	die	die	DEM			Masc	Acc	Sg	St	Subst
den	die	die	DEM			NoGend	Dat	Pl	St	Attr
den	die	die	DEM			Masc	Acc	Sg	St	Attr
den	die	die	ART	Def		Masc	Acc	Sg	St	Subst
den	die	die	ART	Def		NoGend	Dat	Pl	St	Attr
den	die	die	ART	Def		Masc	Acc	Sg	St	Attr
Kinderbänken	Kinderbank	Kind<~>er<#>bank	NN			Fem	Dat	Pl
ausruhen	ausruhen	aus<#>ruh<~>en	V		3			Pl				Subj	Pres
ausruhen	ausruhen	aus<#>ruh<~>en	V		3			Pl				Ind	Pres
ausruhen	ausruhen	aus<#>ruh<~>en	V		1			Pl				Subj	Pres
ausruhen	ausruhen	aus<#>ruh<~>en	V		1			Pl				Ind	Pres
ausruhen	ausruhen	aus<#>ruh<~>en	V								Inf
.	.	.	PUNCT	Period

The transducer can be selected as an argument of option -t:

$ echo "Ihr\nkönnt\neuch\nauf\nden\nKinderbänken\nausruhen\n." | ./dwdsmor.py -E -n -t lib/dwdsmor-root.ca
Wordform	Lemma	Segmentation	Process	Means	POS	Subcategory	Person	Gender	Case	Number	Inflection	Function	Nonfinite	Mood	Tense	Metalinguistic	Characters
Ihr	sie	sie			PPRO	Pers	3	NoGend	Gen	Pl						Old	CAP
Ihr	sie	sie			PPRO	Pers	3	Fem	Dat	Sg							CAP
Ihr	sie	sie			PPRO	Pers	3	Fem	Gen	Sg						Old	CAP
Ihr	Sie	Sie			PPRO	Pers	3	NoGend	Gen	Pl						Old
Ihr	ihr	ihr			PPRO	Pers	2		Nom	Pl							CAP
Ihr	ihre	ihre			POSS			Neut	Acc	Sg	NoInfl	Attr					CAP
Ihr	ihre	ihre			POSS			Neut	Nom	Sg	NoInfl	Attr					CAP
Ihr	ihre	ihre			POSS			Masc	Nom	Sg	NoInfl	Attr					CAP
Ihr	Ihre	Ihre			POSS			Neut	Acc	Sg	NoInfl	Attr
Ihr	Ihre	Ihre			POSS			Neut	Nom	Sg	NoInfl	Attr
Ihr	Ihre	Ihre			POSS			Masc	Nom	Sg	NoInfl	Attr
könnt	können	könn<~>en			V		2			Pl				Ind	Pres
euch	euch	euch			PPRO	Refl	2		Acc	Pl
euch	euch	euch			PPRO	Refl	2		Dat	Pl
euch	ihr	ihr			PPRO	Pers	2		Acc	Pl
euch	ihr	ihr			PPRO	Pers	2		Dat	Pl
auf	auf	auf			ADV
auf	auf	auf			PREP
den	die	die			REL			Masc	Acc	Sg	St	Subst
den	die	die			DEM			Masc	Acc	Sg	St	Subst
den	die	die			DEM			NoGend	Dat	Pl	St	Attr
den	die	die			DEM			Masc	Acc	Sg	St	Attr
den	die	die			ART	Def		Masc	Acc	Sg	St	Subst
den	die	die			ART	Def		NoGend	Dat	Pl	St	Attr
den	die	die			ART	Def		Masc	Acc	Sg	St	Attr
Kinderbänken	Kind + Bank	Kind<+>Bank	COMP	concat	NN			Fem	Dat	Pl
ausruhen	ruhen	ruh<~>en	DER	part(aus)	V		3			Pl				Subj	Pres
ausruhen	ruhen	ruh<~>en	DER	part(aus)	V		3			Pl				Ind	Pres
ausruhen	ruhen	ruh<~>en	DER	part(aus)	V		1			Pl				Subj	Pres
ausruhen	ruhen	ruh<~>en	DER	part(aus)	V		1			Pl				Ind	Pres
ausruhen	ruhen	ruh<~>en	DER	part(aus)	V								Inf
.	.	.			PUNCT	Period

CSV, JSON, and YAML outputs are available with options -c, -j, and -y respectively.

paradigm.py is Python script for the generation of paradigms of lexical words in written German by means of a DWDSmor transducer:

$ ./paradigm.py -h
usage: paradigm.py [-h] [-c] [-C] [-E] [-H] [-i {1,2,3,4,5}] [-I {1,2,3,4,5}] [-j] [-n] [-N]
                   [-o] [-O] [-p {ADJ,ART,CARD,DEM,INDEF,NN,NPROP,POSS,PPRO,REL,V,WPRO}] [-P]
                   [-s] [-S] [-t TRANSDUCER] [-u] [-v] [-y] lemma [output]

positional arguments:
  lemma                 lemma (determiners: Fem Nom Sg; nominalised adjectives: Wk)
  output                output file (default: stdout)

optional arguments:
  -h, --help            show this help message and exit
  -c, --csv             output CSV table
  -C, --force-color     preserve color and formatting when piping output
  -E, --no-empty        do not output empty columns or values
  -H, --no-header       suppress table header
  -i {1,2,3,4,5}, --lemma-index {1,2,3,4,5}
                        homographic lemma index
  -I {1,2,3,4,5}, --paradigm-index {1,2,3,4,5}
                        paradigm index
  -j, --json            output JSON object
  -n, --no-cats         do not output category names
  -N, --no-lemma        do not output lemma, lemma index, paradigm index, and lexical categories
  -o, --old             output also archaic forms
  -O, --oldorth         output also forms in old spelling
  -p {ADJ,ART,CARD,DEM,INDEF,NN,NPROP,ORD,POSS,PPRO,REL,V,WPRO}, --pos {ADJ,ART,CARD,DEM,INDEF,NN,NPROP,ORD,POSS,PPRO,REL,V,WPRO}
                        part of speech
  -P, --plain           suppress color and formatting
  -s, --nonst           output also non-standard forms
  -S, --ch              output also forms in Swiss spelling
  -t TRANSDUCER, --transducer TRANSDUCER
                        transducer file (default: lib/dwdsmor-index.a)
  -u, --user-specified  use only user-specified information
  -v, --version         show program's version number and exit
  -y, --yaml            output YAML document

By default, paradigm.py outputs a similar TSV table as dwdsmor.py:

$ ./paradigm.py -E Bank
Lemma	Lemma Index	POS	Gender	Case	Number	Paradigm Forms
Bank	1	NN	Fem	Nom	Sg	Bank
Bank	1	NN	Fem	Acc	Sg	Bank
Bank	1	NN	Fem	Dat	Sg	Bank
Bank	1	NN	Fem	Gen	Sg	Bank
Bank	1	NN	Fem	Nom	Pl	Bänke
Bank	1	NN	Fem	Acc	Pl	Bänke
Bank	1	NN	Fem	Dat	Pl	Bänken
Bank	1	NN	Fem	Gen	Pl	Bänke
Bank	2	NN	Fem	Nom	Sg	Bank
Bank	2	NN	Fem	Acc	Sg	Bank
Bank	2	NN	Fem	Dat	Sg	Bank
Bank	2	NN	Fem	Gen	Sg	Bank
Bank	2	NN	Fem	Nom	Pl	Banken
Bank	2	NN	Fem	Acc	Pl	Banken
Bank	2	NN	Fem	Dat	Pl	Banken
Bank	2	NN	Fem	Gen	Pl	Banken

For a condensed version, the options -n and -N can be specified. The DWDS homographic lemma index can be selected with option -i:

$ ./paradigm.py -n -N -i 1 Bank
Paradigm Categories	Paradigm Forms
Nom Sg	Bank
Acc Sg	Bank
Dat Sg	Bank
Gen Sg	Bank
Nom Pl	Bänke
Acc Pl	Bänke
Dat Pl	Bänken
Gen Pl	Bänke

The default transducer for paradigm generation is dwdsmor-index.a and restricted to inflection only. Paradigms for word-formation products which are unavailable in the DWDS can be generated with the transducer dwdsmor.a:

$ ./paradigm.py -n -N -t lib/dwdsmor.a Kinderbank
Paradigm Categories	Paradigm Forms
Nom Sg	Kinderbank
Acc Sg	Kinderbank
Dat Sg	Kinderbank
Gen Sg	Kinderbank
Nom Pl	Kinderbanken, Kinderbänke
Acc Pl	Kinderbanken, Kinderbänke
Dat Pl	Kinderbanken, Kinderbänken
Gen Pl	Kinderbanken, Kinderbänke

Note that this transducer does not know of DWDS homographic lemma indices.

Again, options -c, -j, -y select alternative CSV, JSON, and YAML outputs.

Contact

Feel free to contact Andreas Nolda for questions regarding the lexicon or the grammar and Gregor Middell for question related to the integration of DWDSmor into your corpus-annotation pipeline.

Bibliography

• Berlin-Brandenburg Academy of Sciences and Humanities (BBAW) (ed.) (n.d.). DWDS – Digitales Wörterbuch der deutschen Sprache: Das Wortauskunftssystem zur deutschen Sprache in Geschichte und Gegenwart. https://www.dwds.de
• Fitschen, Arne (2004). Ein computerlinguistisches Lexikon als komplexes System. Ph.D. thesis, Universität Stuttgart. PDF
• Nolda, Andreas (2022). Headedness as an epiphenomenon: Case studies on compounding and blending in German. In Headedness and/or Grammatical Anarchy?, ed. by Ulrike Freywald, Horst Simon, and Stefan Müller, Empirically Oriented Theoretical Morphology and Syntax 11, Berlin: Language Science Press, 343–376. PDF.
• Schmid, Helmut (2006). A programming language for finite state transducers. In Finite-State Methods and Natural Language Processing: 5th International Workshop, FSMNLP 2005, Helsinki, Finland, September 1–2, 2005, ed. by Anssi Yli-Jyrä, Lauri Karttunen, and Juhani Karhumäki, Lecture Notes in Artificial Intelligence 4002, Berlin: Springer, 1263–1266. PDF.
• Schmid, Helmut, Arne Fitschen, and Ulrich Heid (2004). SMOR: A German computational morphology covering derivation, composition, and inflection. In LREC 2004: Fourth International Conference on Language Resources and Evaluation, ed. by Maria T. Lino et al., European Language Resources Association, 1263–1266. PDF
• Sennrich, Rico and Beta Kunz (2014). Zmorge: A German morphological lexicon extracted from Wiktionary. In LREC 2014: Ninth International Conference on Language Resources and Evaluation, ed. by Nicoletta Calzolari et al., European Language Resources Association, 1063–1067. PDF.

License

As the original SMOR and SMORLemma grammars, the DWDSmor grammar is licensed under the GNU General Public Licence v2.0. The rest of this project is licensed under the GNU Lesser General Public License v3.0.

Andreas Nolda andreas.nolda@bbaw.de