gCLIMB for German

General

This page provides background information on phenomena and implementations included in gCLIMB for German.

In this document, I and me refers to AntskeFokkens.

Disclaimer

  1. This documentation is under construction and incomplete. At present, the only way to get detailed into implementations of gCLIMB is by looking at the grammars.
  2. I am currently finishing my thesis. In the near future, this page will therefore mostly contain background information that completes work described in my PhD. If you have request on other explanations or documentation, please contact me and I’ll see if I can shift priorities around.
  3. This documentation will be restructured soon

Where to get gCLIMB

For now, e-mail me. The repository should be accessible for guest accounts soon.

Phenomena

This section provides an incomplete list of phenomena implemented in gCLIMB for German. It describes when the phenomenon was added, when revised and what decisions were made concerning the scope of the phenomenon (i.e. what to do rare constructions, examples that are awkward, but not ungrammatical etc.)

Word order for verbs (left and right brackets, relative order in right bracket)

In principle covered since Fokkens (2011).

Old Bug report

In previous versions there was a problem with interaction between polar questions and restricting the verbal cluster. The arg-comp analysis incorrectly accepted the following sentences:

           
Topological fields: Vorfeld Left Bracket Mittelfeld Right Bracket Nachfeld
German: * Hat   schlafen der Mann können  
Transliteration:   have.3rd.sg   sleep.inf the.nom man.nom can.nf  
German: * Hat mich sehen der Mann können  
Transliteration:   have.3rd.sg me.acc see.inf the.nom man.nom can.nf  

The problem was corrected at a later stage. More information can be found in the repository overview below.

Development stages

   
repository version addition/revision
19613 Addition of polar questions, including bug for arg-comp analysis
19617 Bug fixed

Split verbal clusters

Split verbal clusters are a specific form of partial VP fronting, where the main verb is placed in the Vorfeld (optionally accompanied by one or more arguments) and at least one auxiliary is left in the Right Bracket. An example (meaning you should be able to sleep here):

           
Topological fields: Vorfeld Left Bracket Mittelfeld Right Bracket Nachfeld
German: Schlafen solltest Du hier auf jeden Fall können  
Transliteration: sleep.inf should.2nd.sg you.nom here in any case can.inf  

Split verbal clusters are beyond doubt grammatical, but marked and rarely used. gCLIMB allows you to either include or exclude them. These two possibilities have been present in gCLIMB since Fokkens (2011), though revisions have been made.

Old Bug report

Bug 1

At certain stages in the past, the argument-composition analysis in gCLIMB required the fronted main verb to fit subcategorization requirements of the finite verb in the left bracket. This is incorrect, it should fit subcategorization requirements of the auxiliary in the Nachfeld that governs it. The auxiliary+construction analysis in this version is correct. The repository table below provides the information that is currently available about this bug. Where it was introduced exactly and when it was fixed completely is still under investigation.

Subversion version 19613 (2 major updated after the place where the bug was found) largely recovers the bug. Split clusters work in principle, except for cases where the main verb and dative argument are fronted and subject and accusative argument remain in the Mittelfeld. The example below illustrates such a structure (The man should be allowed to give the wine to me):

           
Topological fields: Vorfeld Left Bracket Mittelfeld Right Bracket Nachfeld
German: Mir geben sollte der Mann den Wein dürfen  
Transliteration: me.dat give.inf should.3rd.sg the.nom.sg mann.nom.sg the.acc wine.acc may.inf  

Bug 2

As informal complements were introduced, the interaction between split verbal cluster of the auxiliary + verbal construction analysis and informal vcomps was broken. It accepted the following sentence (ungrammatical unless schläft er? would be direct speech):

*Schläft      er sagt       der        Mann
 sleep.3rd.sg he say.3rd.sg the.nom.sg man.nom.sg

Development stages

   
repository version addition/revision
Older branch Split cluster already optionally included. The aux-rule analysis uses a hack that violates principles of semantic compositionality
19578 Bug on subcategorization requirements for arg-comp analysis identified
19613 Bug on subcategorization partially fixed (no partial VP with only dative argument fronted)
19617 Informal vcomps are added, bug interacting with aux+vconstr split cluster analysis
19732 Bug on subcategorization completely fixed
… …
23282 Bugs known to be fixed completely, more principled analysis for aux-rule and split clusters (Fokkens, in progress)

Subordinate clauses, VCOMP

Supports both subordinates introduced by complementizers dass(that) and ob (whether) with finite verb final word order, as well as informal (spoken language), where the complementizer is dropped and the finite verb is placed in second position.

Er sagte            dass der        Mann        mir     das        Buch        schenken würde.
He say.past.3rd.sg  that the.nom.sg man.nom.sg  me.dat  the.acc.sg book.acc.sg give.inf would.3rd.sg
''He said that the man would give the book to me.''


Er sagte            der        Mann        würde        mir     das        Buch        schenken.
He say.past.3rd.sg  the.nom.sg man.nom.sg  would.3rd-sg me.dat  the.acc.sg book.acc.sg give.inf
''He said that the man would give the book to me.''


*Er sagte            der        Mann        mir     das        Buch        schenken würde.
 He say.past.3rd.sg  the.nom.sg man.nom.sg  me.dat  the.acc.sg book.acc.sg give.inf would.3rd.sg


*Er sagte            dass der        Mann        würde        mir     das        Buch        schenken.
 He say.past.3rd.sg  that the.nom.sg man.nom.sg  would.3rd.sg me.dat  the.acc.sg book.acc.sg give.inf

Old bug

When first introduced, the grammar slightly overgenerated. Matrix verbs selecting a question complement could be combined with a v2 informal complement. Examples like the following were accepted:

*Er fragt      der        Mann       schenkt     mir    das        Buch.
 He ask.3rd.sg the.nom.sg man.nom.sg give.3rd.sg me.dat the.dat.sg book.dat.sg

Details on when the bug was introduced and when it was fixed can be found below.

Development stages

   
repository version addition/revision
19617 subordinates introduced by ‘dass’ and ‘ob’ as well as informal vcomps introduced. Bug: overgenerating informals for verbs that do not allow for this
19769 Bug fixed

Adverbial Negation

Comments:

  • Fronting of the negative adverb in German sounds unnatural, but could be acceptable in poetry. In the initial implementation it is accepted by the grammar (i.e. parsed and generated)
?Nicht schläft           der     Mann.
Not    sleep.pres.3rd.sg the.nom man.nom.
''The Man doesn't sleep.''

Development stages

  •    
    repository version addition/revision
    19613 Initial implementation of nicht with free position of adverbial

Object raising

old bug

When object raising was introduced, the interaction of VP-coordination and object-raising was broken. Examples like the following were not covered.

Er will        tanzen    und kann       lachen.
He want.3rd.sg dance.inf and can.3rd.sg laugh.inf
''He wants to dance and can laugh.''

Er will        schlafen  und tanzt.
He want.3rd.sg sleep.inf and dance.3rd.sg
''He wants to sleep and dances.''

Development stages

   
repository version addition/revision
19732 Implementation of object raising, bug of interaction vp-coord and obj-raising for arg-comp present
… …
23282 Bug known to be fixed

References

Fokkens, Antske (2011) Metagrammar engineering: Towards systematic ex- ploration of implemented grammars. In Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Lan- guage Technologies, pages 1066–1076, Portland, Oregon, USA: Association for Computational Linguistics.

Fokkens, Antske (in progress) Enhancing empirical research for linguistic precision grammars. PhD thesis. Saarland University.
NB References to this work in the documentation are already written. Please contact me if you are interested in this information before the PhD is done.

Last update: 2012-12-31 by AntskeFokkens [edit]