-
Notifications
You must be signed in to change notification settings - Fork 0
/
formgram-gpsg.tex
994 lines (743 loc) · 24.9 KB
/
formgram-gpsg.tex
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
%% -*- coding:utf-8 -*-
\section{Generalized Phrase Structure Grammar (GPSG)}
\subtitle{Generalized Phrase Structure Grammar (GPSG)}
\huberlintitlepage[22pt]
\outline{
\begin{itemize}
\item Introduction and basic terms
\item Phrase structure grammar and \xbar Theory
\item Government \& Binding (GB)
\item \alert{Generalized Phrase Structure Grammar (GPSG)}
\item {Feature descriptions, feature structures and models}
\item Lexical Functional Grammar (LFG)
%\item PATR
\item Categorial Grammar (CG)
\item Head-Driven Phrase Structure Grammar (HPSG)
%\item Konstruktionsgrammatik (CxG)
\item Tree Adjoning Grammar (TAG)
\end{itemize}
%\tableofcontents
}
\frame{
\frametitle{Reading material}
\citew[Chapter~5]{MuellerGT-Eng} without Section~5.1.4 about semantics.
}
\frame{
\frametitle{Generalized Phrase Structure Grammar (GPSG)}
\begin{itemize}
\item GPSG was developed as an answer
to Transformational Grammar at the end of the 1970s.
\pause
\item Main publication: \citet*{GKPS85a}
\pause
\item \citet{Uszkoreit87a} developed large GPSG fragment of German.
\pause
\item Chomsky showed PSGs to be inadequate.\\
GPSG extends PSG in ways that make it possible to address Chomsky's monita:
\begin{itemize}
\item categories may be complex \citep{Harman63a}
\pause
\item different treatment of local reordering
\pause
\item passive as metarule
\pause
\item non-local dependencies as a series of local dependencies
\end{itemize}
\pause
\item We will deal with each of these innovations in what follows.
\end{itemize}
}
\subsection{General remarks on the representational format}
\subsubsection{Categories and \xbar Theory}
\frame[shrink=0]{
\frametitle{General remarks on the representational format}
\begin{itemize}
\item Categories are sets of feature value pairs.
\pause
\item Lexical entries have a feature \subcat. The value is a number
which says something about the kind of grammatical rules in which the word can be used.
\pause
\item Examples from \citew{Uszkoreit87a}:
\medskip
\begin{tabular}{@{}l@{~$\to$~}ll@{}}
V2 & H[5] & (kommen `come', schlafen `sleep')\\
V2 & H[6], N2[Case Acc] & (kennen `know', suchen `search')\\
V2 & H[7], N2[Case Dat] & (helfen `help', vertrauen `trust')\\
V2 & H[8], N2[Case Dat], N2[Case Acc] & (geben `give', zeigen `show')\\
V2 & H[9], V3[+dass] & (wissen `know', glauben `believe')\\
\end{tabular}
\medskip
These rules license VPs: the combination verb \& complements, but not subject.
\pause
\item The numbers following the category symbols (V or N) indicate the
\xbar~level.\\
The maximum level of a verbal projection is three rather than two.
\pause
\item H stands for Head.
\end{itemize}
}
\subsubsection{Principles: The Head Feature Convention}
\frame{
\frametitle{Principles: The Head Feature Convention}
Head Feature Convention:\\
The mother node and the head daughter must bear the same head features unless indicated otherwise.
}
\subsubsection{Metarules and ID/LP format}
\frame{
\frametitle{Metarules and ID/LP format}
Two further innovations of GPSG:
\begin{itemize}
\item Metarules: Additional phrase structure rules are licensed via metarules.
\item ID/LP format: Constraints on linearization are separated from immediate dominance.
\end{itemize}
These two tools will be discussed with respect to our set of phenomena.
}
\subsection{Local reordering \& Verb position}
\frame{
\frametitle{Local reordering}
\begin{itemize}
\item Arguments can appear in almost any order in the German \mf.
\eal
\ex
\gll {}[weil] \gruen{der} \gruen{Mann} \rot{dem} \rot{Kind} \blau{das} \blau{Buch} gibt\\
\spacebr{}because the.\NOM{} man the.\DAT{} child the.\ACC{} book gives\\
\glt `because the man gives the book to the child'
\ex
\gll {}[weil] \gruen{der} \gruen{Mann} \blau{das} \blau{Buch} \rot{dem} \rot{Kind} gibt\\
\spacebr{}because the.\NOM{} man the.\ACC{} book the.\DAT{} child gives\\
\ex\label{ex-das-buch-der-mann-der-frau-gibt}
\gll {}[weil] \blau{das} \blau{Buch} \gruen{der} \gruen{Mann} \rot{dem} \rot{Kind} gibt\\
\spacebr{}because the.\ACC{} book the.\NOM{} man the.\DAT{} child gives\\
\ex
\gll {}[weil] \blau{das} \blau{Buch} \rot{dem} \rot{Kind} \gruen{der} \gruen{Mann} gibt\\
\spacebr{}because the.\ACC{} book the.\DAT{} child the.\NOM{} man gives\\
\ex
\gll {}[weil] \rot{dem} \rot{Kind} \gruen{der} \gruen{Mann} \blau{das} \blau{Buch} gibt\\
\spacebr{}because the.\DAT{} child the.\NOM{} man the.\ACC{} book gives\\
\ex
\gll {}[weil] \rot{dem} \rot{Kind} \blau{das} \blau{Buch} \gruen{der} \gruen{Mann} gibt\\
\spacebr{}because the.\DAT{} child the.\ACC{} book the.\NOM{} man gives\\
\zl
\end{itemize}
}
\frame{
\frametitle{Motivation for linearization rules (I)}
%\smallframe
\savespace
Motivation: Permutation with phrase structure rules $\to$\\
we need six phrase structure rules for ditransitive verbs in verb-final position:
\ea
\begin{tabular}[t]{@{}l@{ }l@{ }l@{ }l@{ }l@{ }}
S & $\to$ NP[nom]& NP[dat] & NP[acc] & V\\
S & $\to$ NP[nom]& NP[acc] & NP[dat] & V\\
S & $\to$ NP[acc]& NP[nom] & NP[dat] & V\\
S & $\to$ NP[acc]& NP[dat] & NP[nom] & V\\
S & $\to$ NP[dat]& NP[nom] & NP[acc] & V\\
S & $\to$ NP[dat]& NP[acc] & NP[nom] & V\\
\end{tabular}
\z
}
\frame{
\frametitle{Motivation for linearization rules (II)}
Plus six rules for verb-initial position:
\ea
\begin{tabular}[t]{@{}l@{ }l@{ }l@{ }l@{ }l}
S & $\to$ V NP[nom]& NP[dat] & NP[acc]\\
S & $\to$ V NP[nom]& NP[acc] & NP[dat]\\
S & $\to$ V NP[acc]& NP[nom] & NP[dat]\\
S & $\to$ V NP[acc]& NP[dat] & NP[nom]\\
S & $\to$ V NP[dat]& NP[nom] & NP[acc]\\
S & $\to$ V NP[dat]& NP[acc] & NP[nom]\\
\end{tabular}
\z
A generalization is missed.
Similarly for transitive verbs and other valence frames.
}
\frame{
\frametitle{Abstraction from linear order: Dominance}
\begin{itemize}
\item \citet*{GKPS85a}:\\
Separation of \alert{immediate dominance} = ID and \alert{linear precedence} = LP.
\pause
\item Dominance rules do not constrain the order of the daughters.
\ea
\begin{tabular}[t]{@{}l@{ }l}
S & $\to$ V, NP[nom], NP[acc], NP[dat]\\
\end{tabular}
\z
The only thing (\mex{0}) says is that S dominates the other nodes:
\medskip
\centerline{%
\begin{forest}
sm edges
[S
[V]
[{NP[nom]}]
[{NP[dat]}]
[{NP[acc]}] ]
\end{forest}}
\medskip
\pause
\item Since there are no constraints on the order of the elments of the right-hand side, we need one
rule rather than twelve:
\end{itemize}
}
\frame{
\frametitle{Abstraction from linear order: Linear order}
\begin{itemize}
\item LP rules hold for local trees, that is, trees of depth one:
\begin{table}[H]
\centerline{%
\begin{forest}
sm edges
[S
[V]
[{NP[nom]}]
[{NP[dat]}]
[{NP[acc]}] ]
\end{forest}}
\end{table}
~\medskip
$\to$ We can say something about order of V, NP[nom], NP[dat] and NP[acc].
\pause
An LP constraint holds for the whole grammar.\\
If we claim that NP[nom] precedes NP[acc],\\
this holds for rules for strictly transitive verbs as well as for rules for ditransitive verbs.
\end{itemize}
}
\frame{
\frametitle{Getting more restrictive again}
\begin{itemize}
\item Without restriction for the order $\to$ too much freedom
\medskip
\begin{tabular}[t]{@{}l@{ }l}
S & $\to$ V, NP[nom], NP[dat], NP[acc]\\
\end{tabular}
\medskip
The rule admits the following order:
\ea[*]{
\gll \rot{Dem} \rot{Kind} \gruen{der} \gruen{Mann} gibt ein Buch.\\
the.\DAT{} child the.\NOM{} man gives the.\ACC{} book\\
}
\z
\pause
\item Linearization rules rule out such orders.
\ea
\begin{tabular}[t]{@{}l@{~$<$~}l@{}}
V[+MC] & X\\
X & V[$-$MC]\\
\end{tabular}
\z
\textsc{mc} stand for \emph{main clause}.
LP rule states: verb must be placed before all other constituents in main clauses
(+\textsc{mc}) and after all other constituents in dependent clauses ($-$\textsc{mc}).
\end{itemize}
}
\subsection{Passive}
\subsubsection{Passive pre-theoretically}
\frame[shrink=15]{
\frametitle{Passive pre-theoretically (I)}
\savespace\smallexamples
German passive theory-neutrally:
\begin{itemize}
\item The subject is suppressed.
\item If there is an accusative object, this becomes the subject.
\end{itemize}
This holds for all verb classes forming a passive. Independent of the arity of the verb:
\eal
\label{beispiel-arbeiten}
\ex
\gll weil er noch gearbeitet hat\\
because he.\nom{} still worked has\\
\glt 'because he has still worked'
\ex
\gll weil noch gearbeitet wurde\\
because still worked was\\
\glt `because there was still working there'
\zl
\pause
\eal
\label{beispiel-denken}
\ex
\gll weil er an Maria gedacht hat\\
because he.\nom{} on Maria thought has\\
\glt `because he thought of Maria'
\ex
\gll weil an Maria gedacht wurde\\
because on Maria thought was\\
\glt `because Maria was thought of'
\zl
}
\frame[shrink=15]{
\frametitle{Passive pre-theoretically (II)}
\savespace\smallexamples
German passive theory-neutrally:
\begin{itemize}
\item The subject is suppressed.
\item If there is an accusative object, this becomes the subject.
\end{itemize}
\eal
\ex
\gll weil Judit den Weltmeister geschlagen hat\\
because Judit.\nom{} the.\acc{} world.champion beaten has\\
\glt `because Judit has beaten the world champion'
\ex
\gll weil der Weltmeister geschlagen wurde\\
because the.\nom{} world.champion beaten was\\
\glt `because the world champion was beaten'
\zl
\pause
\eal
\ex
\gll weil er ihm den Aufsatz gegeben hat\\
because he.\nom{} him.\dat{} the.\acc{} essay given has\\
\glt `because he has given him the essay'
\ex
\gll weil ihm der Aufsatz gegeben wurde\\
because him.\dat{} the.\nom{} essay given was\\
\glt `because he was given the essay'
\zl
}
\subsubsection{Metarules}
\frame{
\frametitle{Passive and phrase structure grammars}
\begin{itemize}
\item One would have to write down two rules for every active/passive pair in PSG.
\pause
\item GPSG is a non-transformational theory.
\pause
\item Metarule derives passive rules from active rules.
\pause
\medskip
\item These are explained with respect to the subject introduction metarule.
\end{itemize}
}
\frame[shrink=5]{
\frametitle{Introduction of the subject via a metarule (I)}
Our rules look like this:
\ea
\begin{tabular}[t]{@{}l@{~$\to$~}ll@{}}
V2 & H[7], N2[Case Dat] & (helfen `help', vertrauen `trust')\\
V2 & H[8], N2[Case Dat], N2[Case Acc] & (geben `give', zeigen `show')\\
\end{tabular}
\z
\pause
The rules in (\mex{0}) can be used to analyze VPs but not sentences with subject.
\pause
We use a metarule saying: ``If there is a rule of the form `V2 consists of something', then there is
also a rule stating `V3 consists of whatever V2 consists of + an NP in the nominative'{}''.
\pause
Formally:
\ea
V2 $\to$ W $\mapsto$\\
V3 $\to$ W, N2[Case Nom]
\z
W stands for an arbitrary number of categories (whatever).
}
\frame{
\frametitle{Introduction of the subject via a metarule (II)}
\ea
V2 $\to$ W $\mapsto$\\
V3 $\to$ W, N2[Case Nom]
\z
\pause
This metarule takes the rules in (\mex{1}) as input and produces the rules in (\mex{2}):
\ea
\oneline{%
\begin{tabular}[t]{@{}l@{~$\to$~}ll@{}}
V2 & H[7], N2[Case Dat] & (helfen `help', vertrauen `trust')\\
V2 & H[8], N2[Case Dat], N2[Case Acc] & (geben `give', zeigen `show')\\
\end{tabular}
}
\z
\ea
\begin{tabular}[t]{@{}l@{~$\to$~}l@{}}
V3 & H[7], N2[Case Dat], N2[Case Nom] \\
V3 & H[8], N2[Case Dat], N2[Case Acc], N2[Case Nom] \\
\end{tabular}
\z
\pause
Subject and other arguments are on the same right-hand side of a rule and hence can be permuted,
provided no LP rule is violated.
}
\subsubsection{Passive as metarule}
\frame[shrink=5]{
\frametitle{Passive as metarule}
\begin{itemize}
\item For each active rule with subject and accusative object, a passive rule will be licensed with
the subject suppressed. The relation between the rules is captured.
\pause
\item Differences between Transformational Grammar/GB and GPSG:\\
It is not the case that there are several trees that are related to each other,\\
but rather active rules are related to passive rules.
The active and passive rules can be used to derive two structures independently:\\
(\mex{1}b) is not derived from (\mex{1}a).
\eal
\ex
\gll weil Judit den Weltmeister geschlagen hat\\
because Judit.\nom{} the.\acc{} world.champion beaten has\\
\glt `because Judit has beaten the world champion'
\ex
\gll weil der Weltmeister geschlagen wurde\\
because the.\nom{} world.champion beaten was\\
\glt `because the world champion was beaten'
\zl
The generalization regarding active/passive alternations is captured nevertheless.
\end{itemize}
}
\frame{
\frametitle{Passive in English}
\citet*{GKPS85a} suggest the following metarule:
\ea
VP $\to$ W, NP $\mapsto$\\
VP[PAS] $\to$ W, (PP[\emph{by}])
\z
This rule says that verbs selecting an object can be realized without this object in a passive VP.
Optionally a \emph{by} PP may appear in passive VPs.
(VP corresponds to V2)
}
\frame{
\frametitle{Problems of the passive metarule operating on VP}
\begin{enumerate}
\item Rule does not refer to the type of the verb (not all verbs have a passive).
\pause
% \item Es ist unklar, wie die Semantik parallel zur Syntax aufgebaut werden soll:
% Die Regel in (\mex{0}) unterdrückt ein NP-Argument in der VP.
% Dieses Argument ist aber ein Objekt. Rein syntaktisch stellt das System der GPSG-Metaregeln
% die richtige Menge von ID-Regeln zur Verfügung, in der Semantik muss man aber sicherstellen,
% dass das durch eine Metaregel eingeführte Subjekt mit dem unterdrückten Objekt übereinstimmt.
% \pause
\item Impersonal passive cannot be derived by suppressing an object.
\ea
\begin{tabular}[t]{@{}l@{\hspace{2cm}}l@{}}
V2 $\to$ H[5] & (\emph{arbeiten} `work')\\
V2 $\to$ H[13], PP[\emph{an}] & (\emph{denken} `think')\\
\end{tabular}
\z
So, if the analysis of the passive in English is not revised,\\
the analyses of the passive in English and German will differ.
\pause
\item The German passive metarule could apply to rules including the subject.
\end{enumerate}
}
\subsection{Long"=distance dependencies}
\frame[shrink=5]{
\frametitlefit{Long"=distance dependencies as the result of local dependencies}
\begin{itemize}
\item Until now: verb-initial and verb-final placement of the verb:
\eal
\ex
\gll {}[dass] der Mann dem Kind das Buch \alert{gibt}\\
{}\spacebr{}that the.\NOM{} man the.\DAT{} child the.\ACC{} book gives\\
%\glt `that the man gives the book to the child'
\ex
\gll \alert{Gibt} der Mann dem Kind das Buch?\\
gives the.\NOM{} man the.\DAT{} child the.\ACC{} book\\
%\glt `Does the man give the book to the child?'
\zl
\pause
\item What about verb second placement:
\eal
\ex
\gll Der Mann \alert{gibt} dem Kind das Buch.\\
the.\NOM{} man gives the.\DAT{} child the.\ACC{} book\\
%\glt `The man gives the child the book.'
\ex
\gll Dem Kind \alert{gibt} der Mann das Buch.\\
the.\DAT{} child gives the.\NOM{} man the.\ACC{} book\\
%\glt `The man gives the child the book.'
\zl
\pause
\item V2 is analyzed as a nonlocal dependency via a sequence of local dependencies.
One of the main innovations of GPSG:\\
transformationless analysis of nonlocal dependencies (but also \citew{Harman63a}).
\end{itemize}
}
\subsubsection{Metarules for the introduction of nonlocal dependencies}
\frame{
\frametitle{Metarules for the introduction of nonlocal dependencies}
We take an arbitrary category X out of the set of categories on the right-hand side of the rule and
represent it on the left-hand side after a slash (`/'):
\ea
V3 $\to$ W, \gruen{X} $\mapsto$\\
V3/\gruen{X} $\to$ W
\z
\pause
Given the input in (\mex{1}), the rule creates the rules in (\mex{2}):
\ea
\begin{tabular}[t]{@{}l@{~$\to$~}l@{}}
V3 & H[8], N2[Case Dat], N2[Case Acc], N2[Case Nom]
\end{tabular}
\z
\ea
\begin{tabular}[t]{@{}l@{~$\to$~}l@{}}
V3/N2[Case Nom] & H[8], N2[Case Dat], N2[Case Acc]\\
V3/N2[Case Dat] & H[8], N2[Case Acc], N2[Case Nom]\\
V3/N2[Case Acc] & H[8], N2[Case Dat], N2[Case Nom]\\
\end{tabular}
\z
}
\subsubsection{Rule for binding off nonlocal dependencies}
\frame{
\frametitle{Rule for binding off nonlocal dependencies}
\ea
V3[+Fin] $\to$ X[+Top], V3[+MC]/X
\z
X stands for arbitrary category marked as missing in V3 by `/'.
\pause
Example instantiations of the rule are given in (\mex{1}):
\ea
\begin{tabular}[t]{@{}l@{~$\to$~}l@{~}l@{}}
V3[+Fin] & N2[+Top, Case Nom], & V3[+MC]/N2[Case Nom]\\
V3[+Fin] & N2[+Top, Case Dat], & V3[+MC]/N2[Case Dat]\\
V3[+Fin] & N2[+Top, Case Acc], & V3[+MC]/N2[Case Acc]\\
\end{tabular}
\z
\pause
LP rule: X in (\mex{-1}) is serialized left of anything else
(\eg V3), since it is [+Top].
\ea
{}[+Top] $<$ X
\z
}
\subsubsection{An example analysis}
\frame{
\frametitle{An example analysis}
\centerline{%
\scalebox{0.85}{
\begin{forest}
sm edges
% [{V3[+\textsc{fin}, $+$\textsc{mc}]}
% [\alert<beamer:4>{N2[dat,+\textsc{top}]} [dem Kind;the child,roof] ]
% [{V3[+\textsc{mc}]/\alert<beamer:4>{N2[dat]}}, alert on=<beamer:2>
% [\alert<3>{V[8,+\textsc{mc}]},alert on=<beamer:3> [gibt;gives] ]
% [{N2[nom]} [er;he] ]
% [{N2[acc]} [das Buch;the book, roof] ] ] ]
[{V3[+\textsc{fin}, $+$\textsc{mc}]},s sep+=1em
[{\gruen<4>{N2[dat},+\textsc{top}\gruen<4>{]}} [dem Kind;the child,roof] ]
[\gruen<2>{V3[+\textsc{mc}]/\gruen<1,4>{N2[dat]}}
[\gruen<2>{\gruen<3>{V[}8,\gruen<3>{+\textsc{mc}]}} [{gibt};gives] ]
[\gruen<2>{N2[nom]} [{er};he] ]
[\gruen<2>{N2[acc]} [{das Buch};the book, roof] ] ] ]
\end{forest}
}}
\begin{itemize}
\item Metarule licenses rule introducing dative object into \slasch.
\pause
\item This rule is applied and licenses the subtree for \emph{gibt er das Buch}.
\pause
\item The linearization rule orders the verb left of other constituents (V[+MC] $<$ X).
\pause
\item The constituent following the slash is bound off in the last step.
\end{itemize}
}
\subsubsection{An example with nonlocal dependencies}
\frame{%[shrink=20]{
\frametitle{An example with nonlocal dependencies (I)}
%\savespace
All NPs in (\mex{1}) depend on the same verb:
\ea
\gll Dem Kind gibt er das Buch.\\
the.\dat{} child gives he.\nom{} the.\acc{} book\\
\glt `He gives the child the book.'
\z
Complicated system of linearization rules $\to$ analyze (\mex{0}) with a flat structure.
\pause
But this would not work for:
\ea \gll Wen$_i$ glaubst du, daß ich \_$_i$ gesehen habe?\footnotemark\\
who believe you that I {} seen have\\
\footnotetext{%
\citew[\page84]{Scherpenisse86a}.
}
\glt `Who do you think I saw?'
%\ex
% {\raggedright
% \gll {}[Gegen ihn]$_i$ falle es den Republikanern hingegen schwerer, [~[~Angriffe~\_$_i$] zu lancieren].\footnotemark\\
% {}\spacebr{}against him fall it the Republicans however more.difficult
% \hspaceThis{[~[~}attacks to launch\\
% \par}
% \footnotetext{%
% taz, 08.02.2008, p.\,9.
% }
% \glt `It is, however, more difficult for the Republicans to launch attacks against him.'
\z
(\mex{0}) cannot be explained by local reordering since \emph{wen} does not depend on \emph{glaubst}
but on \emph{gesehen} and \emph{gesehen} is located in a different local subtree.
}
\frame{
\frametitle{An example with nonlocal dependencies (II)}
\begin{itemize}
\item (\mex{1}) is analyzed in several steps: introduction, percolation and finally binding off of information about the long"=distance dependency
\ea
\gll Wen glaubst du, daß ich gesehen habe?\\
who believe you that I seen have\\
\z
\pause
\item \emph{ich gesehen habe} is V3/NP[acc]\\
(grammar rule licensed by a metarule)
\pause
\item \emph{dass ich gesehen habe} is V3/NP[acc]\\
(percolation of \slasch information)
\pause
\item \emph{glaubst du, dass ich gesehen habe} is V3/NP[acc]\\
(percolation of \slasch information)
\pause
\item \emph{Wen glaubst du, dass ich gesehen habe} is V3\\
(binding off of \slasch information in grammar rule)
\end{itemize}
}
\frame{
\frametitle{An example with nonlocal dependencies (III)}
\vfill
\centerline{%
\scalebox{0.75}{%
\begin{forest}
sm edges,empty nodes
[{V3[+\textsc{fin},+\textsc{mc}]}
[{\gruen<3>{N2[acc},+\textsc{top}\gruen<3>{]}} [wen;who] ]
[{V3[+\textsc{mc}]/\gruen<2-3>{N2[acc]}}
[{V[9,+\textsc{mc}]} [glaubst;believes] ]
[{N2[nom]} [du;you] ]
[{V3[+dass,$-$\textsc{mc}]/\gruen<2>{N2[acc]}}
[{}[dass;that] ]
[{V3[$-$dass,$-$\textsc{mc}]/\gruen<1>{N2[acc]}}
[{N2[nom]} [ich;I] ]
[{V[6,$-$\textsc{mc}]} [gesehen habe;seen have,roof] ] ] ] ] ]
\end{forest}
}
}
\vfill
Simplifying assumption: \emph{gesehen habe} behaves like a simplex transitive verb.
\vfill
}
\subsection{Summary and Classification}
\subsubsection{Highlights}
\frame{
\frametitle{Highlights: Across the Board Extraction}
\begin{itemize}
\item Gazdar's \citeyearpar{Gazdar81a} \slasch-based analysis can account for so-called Across the Board extraction \citep{Ross67a}:
\eal\settowidth\jamwidth{(= S/NP \& S/NP)}
\label{ex-atb-gazdar}
\ex[]{ The kennel which Mary made and Fido sleeps in has been stolen. \jambox{(= S/NP \& S/NP)}
}
\ex[]{ The kennel in which Mary keeps drugs and Fido sleeps has been stolen. \jambox{(= S/PP \& S/PP)}
}
\ex[*]{The kennel (in) which Mary made and Fido sleeps has been stolen. \jambox{(= S/NP \& S/PP)}
}
\zl
Conjuncts have to have the same element in \slasch and this information is percolated further and
then bound off.
\pause
\item Such sentences are a miracle for transformational analyses:\\
Why must two transformations move something of the same category?\\
How can two different things land in the same position?
\end{itemize}
}
\subsubsection{Problems}
\frame{
\frametitle{Problems}
\begin{itemize}
\item representation of valence and morphology
\item partial fronting
\item generative capacity
\end{itemize}
}
\subsubsubsection{Representation of valence and morphology}
\frame{
\frametitle{Representation of valence and morphology}
\begin{itemize}
\item Morphology has to access valence information:
\eal\settowidth\jamwidth{(nominative, accusative, PP[mit])}
\ex[]{
\gll lös-bar\\
solv-able\\ \jambox{(nominative, accusative)}
}
\ex[]{
\gll vergleich-bar\\
compar-able\\ \jambox{(nominative, accusative, PP[mit])}
}
\ex[*]{
\gll schlaf-bar\\
sleep-able\\ \jambox{(nominative)}
}
\ex[*]{
\gll helf-bar\\
help-able\\\jambox{(nominative, dative)}
}
\zl
\item Generalization: \emph{bar} adjectives can be formed from verbs governing an accusative.
\pause
\item This information is inaccessable in GPSG. Only valence numbers and this number does
not even tell us whether there is an accusative. There may be a bunch of different rules
(active/passive) with or without the accusative.
\pause
\item Valence must contain detailed descriptions of arguments (CG, LFG, HPSG).
\end{itemize}
}
\subsubsubsection{Partial fronting (I)}
\frame{
\frametitle{Partial fronting}
German allows the fronting of (partial) VPs:
\eal
\ex
\gll [Erzählen] wird \gruen{er} \rot{seiner} \rot{Tochter} \blau{ein} \blau{Märchen} können.\\
\spacebr{}tell will he.\NOM{} his.\DAT{} daughter a.\ACC{} fairy.tale can\\
\glt `He will be able to tell his daughter a fairy tale.'
\ex
\gll [\blau{Ein} \blau{Märchen} erzählen] wird \gruen{er} \rot{seiner} \rot{Tochter} können.\\
\spacebr{}a.\ACC{} fairy.tale tell will he.\NOM{} his.\ACC{} daughter can\\
\ex
\gll [\rot{Seiner} \rot{Tochter} \blau{ein} \blau{Märchen} erzählen] wird \gruen{er} können.\\
\spacebr{}his.\DAT{} daughter a.\ACC{} fairy.tale tell will he.\NOM{} can\\
\zl
Arguments not realized in the fronted VP have to be realized in the \mf.
}
\frame[shrink=7]{
\frametitle{Partial fronting (II)}
\begin{itemize}
\item Arguments missing in initial position have to be realized in the \mf.\\
The case in the \mf has to match the requirement of the verb in the \vf:
\eal
\ex[]{
\gll Verschlungen hat er es nicht.\\
devoured has he.\nom{} it.\acc{} not\\
\glt `He did not devour it.'
}
\ex[*]{
\gll Verschlungen hat er nicht.\\
devoured has he.\nom{} not\\
}
\ex[*]{
\gll Verschlungen hat er ihm nicht.\\
devoured has he.\nom{} him.\dat{} not\\
}
\zl
\pause
\item But this is impossible to do with the standard treatment of valence in GPSG.
\pause
\item Combinations of verbs with arguments are licensed by PSG rules referring to numbers.
\item But the objects can only be missing when they are realized in the \mf.\\
How is this connection established?
\pause
\item \citet{Nerbonne86a} and \citet{Johnson86a}: different representation of valence.\\
One similar to Categorial Grammar.
\end{itemize}
}
\frame{
\frametitle{Generative capacity}
\begin{itemize}
\item The generative capacity of GPSG corresponds to those of context free grammars.
\pause
\item Being restrictive was one of the goals of GPSG.
\pause
\item But \citet{Shieber85a} and \citet{Culy85a}:\\
there are languages that cannot be described with context free grammars.\\
(see also \citew{Pullum86a} for historical remarks)
\pause
\item This means that GPSG is not powerful enough to describe all languages.
\pause
\bigskip
\item All mentioned problems are fixed in HPSG, the successor of GPSG.
\end{itemize}
}
% <!-- Local IspellDict: en_US-w_accents -->