File size: 76,389 Bytes
6fa4bc9
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
{
    "paper_id": "W89-0108",
    "header": {
        "generated_with": "S2ORC 1.0.0",
        "date_generated": "2023-01-19T03:44:50.806412Z"
    },
    "title": "Criteria for Computational Models of Morphology: The Two-Level M odel as an NLP Framework",
    "authors": [
        {
            "first": "Jordan",
            "middle": [],
            "last": "Zlatev",
            "suffix": "",
            "affiliation": {},
            "email": ""
        },
        {
            "first": "A",
            "middle": [],
            "last": "Bstract",
            "suffix": "",
            "affiliation": {},
            "email": ""
        }
    ],
    "year": "",
    "venue": null,
    "identifiers": {},
    "abstract": "Computational models of morphology are best seen not as morphologi cal models but rather ets natural language processing frameworks which can express descriptions in the style of one morphological model or the other, and even go further, but without necessarily being bound by \"purely\" theoretical considerations. Criteria for their adequacy can be derived by treating them (together with the linguistic descriptions that are expressed in their formalisms) as NLP systems, for which a number of goals can be stated, among which are sufficient coverage, efficiency, augmentability and flexibility. The two-level model (T W O L) of Kimmo Koskenniemi is the main object o f attention in this article and examples of its applicability to Bulgarian morphology are presented.",
    "pdf_parse": {
        "paper_id": "W89-0108",
        "_pdf_hash": "",
        "abstract": [
            {
                "text": "Computational models of morphology are best seen not as morphologi cal models but rather ets natural language processing frameworks which can express descriptions in the style of one morphological model or the other, and even go further, but without necessarily being bound by \"purely\" theoretical considerations. Criteria for their adequacy can be derived by treating them (together with the linguistic descriptions that are expressed in their formalisms) as NLP systems, for which a number of goals can be stated, among which are sufficient coverage, efficiency, augmentability and flexibility. The two-level model (T W O L) of Kimmo Koskenniemi is the main object o f attention in this article and examples of its applicability to Bulgarian morphology are presented.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Abstract",
                "sec_num": null
            }
        ],
        "body_text": [
            {
                "text": "T h e cr ite r ia fo r w h a t a \" m o r p h o lo g ic a l m o d e l\" sh ou ld b e a b le t o a c c o u n t for, a n d th e m a n n er in w h ich th is s h o u ld b e d o n e , h ave risen h igh d u rin g th e past fe w y ea rs in a c c o r d a n c e w ith th e situ a tio n in th e n e ig h b o u rin g lin gu istic \" levels\" o f s y n ta x a n d sem a n tics. A p a r t fr o m th e tr a d itio n a l req u irem en ts fo r lin gu istic fe l ic it y ( \" c a p tu r in g th e g e n e ra liz a tio n s\" ), rig ou r, a 1988, K a t a ja a n d K o s k e n n ie m i 1988, C a ld e r 1 9 8 9 ). B u t ev en th e se c a n w ith o u t p e rtie s , has a lso th e b e s t ch a n ce s o f b e in g th e o r e tic a lly sig n ifica n t as w ell, th o u g h th is is a so m e w h a t co n tro v e rsia l m a tte r. A v ie w p o in t th a t is a t lea st le s s co n tro v e rcia l is th a t th e g o a ls o f c o m p u ta tio n a l a n d th e o r e tic a l lin g u istics d iffer.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Introduction",
                "sec_num": "1"
            },
            {
                "text": "S h ieb er (1 9 8 7 ), fo r e x a m p le , has c la im e d th a t th ese d iffe re n ces, e s p e c ia lly c o n cern in g \" restrictiv en ess\" , are s o essen tia l th a t fr o m a c o m p u ta tio n a l p e r s p e c tiv e o n e is m o re in terested in w h a t th e lin g u istic th e o rie s sa y th a n h o w th e y sa y it and th a t it is m ea n in g fu l t o tr y t o se p a ra te th e o rie s ( \" h o w \" ) fr o m th eir a n a ly ses ( \" w h a t\" ) a n d co n c e n tr a te o n th e la tte r in c o m p u ta tio n a l m o d e ls. ",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Introduction",
                "sec_num": "1"
            },
            {
                "text": "I b e liev e th a t o n e c o u ld sa y th a t th e a im s o f c o m p u ta tio n a l a n d th e o r e tic a l lin gu istics e v en tu a lly co n v e rg e , n a m e ly t o g a in a b e tte r u n d e rsta n d in g o f th e n atu re o f h um an la n g u a g e a n d o f its user. S till th e y d iffer in th eir m e th o d s . la n g u a g e p r o c e s s in g a t a su fficien tly th e o r e tic a l level. O n th e o th e r h a n d it is th e o r e tic a l lin g u istics th a t sh o u ld s ta n d fo r th e \" c o n c e p tu a l in sigh ts\" , th e new id e a s a n d th e q u e st fo r lin g u istic u n iversals. O f co u rse , th e clo s e r th e co n n e c tio n b e tw e e n c o m p u ta tio n a l a n d th e o r e tic a l lin gu istics, th e b e tt e r , b u t at least to b e g in w ith , th is is n o t a necessity. W h a t I 'm a im in g a t is t o say th a t co m p u ta tio n a l a n w h e th e r it is \" g e n e ra l\" a n d in w h a t sen se, I w ill c o m e to th a t la ter. R ig h t n ow an im p o r ta n t (in m y o p in io n ) q u e stio n arises, n a m ely: N o w w h a t k in d o f a n im a l is th a t? T h is n e ed h a rd ly b e d efin ed fo r \" in siders\" , b u t fo r s o m e o n e u n fa m ilia r w ith th e ja r g o n in th e field, it sh ou ld b e e n o u g h to sa y th a t an N L P sy ste m ca n b e re g a rd e d as a u n ity o f (a t lea st) th e follo w in g ele m en ts: (1 ) a n im p le m e n ta b le fo rm a lis m , (2 ) a p ro ce s s in g m ech a n ism , a n d 3lin g u is tic k n o w le d g e ex p ressed in th e form a lism . 1 o g y o f a n u m b er o f la n g u a g es ra n g in g fr o m F in n ish (1 9 8 3 ) t o J a p a n e se (A la m 1 983) a n d O ld C h u rch S la v on ic (L in d s te d t 1 9 8 6 ). N o w w h ile th is im p lie s th a t th e T W O L -fr a m e w o r k is g en era l in th e sense th a t it has a p o t e n t i a l l y la rg e co v e ra g e , it d o e s n o t m ea n th a t T W O L is \" g en era l\" in th e sen se th a t it ca n b e a p p lied t o all o f th e w o r ld 's la n g u a g es a n d th eir m o r p h o lo g y -in fle ctio n a l a LEXICON o-a-ov-e/P ozi # \"MASC SING\"; azi # \"FEM SING\"; ova # \"NEUTR SING\" ezi # \"PLUR\" , i.e. o v e rg e n e ra tio n . T h e rea son is th a t i f a fin ite -sta te m ech a n ism allow s A B a n d B C , th en it m u st a ls o a llo w A B C , w h ich in th is ca se w e w a n t t o fo r b id .",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Computational Models as NLP Frameworks",
                "sec_num": "2"
            },
            {
                "text": "S im ilar p ro b le m s w ith th e E n g lish p refix u n -, a re d iscu sse d in (K a r ttu n e n a n d W it te n b u r g 1 983). is th e c o n te x t w h ich ch a ra cterises all B u lg a ria n in te rro g a tiv e s, fo llo w e d b y th e \" rela tiv e sig n \" ):",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Computational Models as NLP Frameworks",
                "sec_num": "2"
            },
            {
                "text": "( (3) \"Palatalization of Velars\" Cx: C y <=> _(V:0) FrontV: whe r e Cx in (k g x) Cy in (c z s) m a t c h e d ;",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Computational Models as NLP Frameworks",
                "sec_num": "2"
            },
            {
                "text": "2) 0 /<= _ k V C (V) (C) # B o th (1 ) an d (2 ) s h o u ld h a v",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Computational Models as NLP Frameworks",
                "sec_num": "2"
            },
            {
                "text": "T h e re a d in g is (fo r o n e u n fa m ilia r w ith th e n o ta tio n ): \" le x ica l k is rea lized as s u rfa ce c ; le x ic a l g as su rfa ce z a n d le x ica l x as su rfa ce s, i f a n d o n ly i f th ey a re fo llo w e d b y a n o p tio n a l le x ica l v o w e l w h ich is rea lized as n o th in g (b e ca u se o f a n o th e r ru le ), a n d a le x ic a l fr o n t v o w e l (rea lized as a n y th in g o n th e su rface i.e. th a t th e y axe T u rin g eq u iv a len t, d o e s n ' t m ea n th a t th e y a re fu n c tio n a lly a n d is m o r e p la u sib le th a n e .g . g e n e ra tiv e p h o n o lo g y .",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": ".A u gm en tability",
                "sec_num": "3"
            },
            {
                "text": "In th e p r e c e d in g I h a v e p re se n ted in a som e w h a t s p e cu la tiv e w a y a n u m b er o f th in k th a t o n e c o u ld sa y th a t it is n o t \" lin g u istic fe lic ity \" . W h a t is m ea n t b y th is is as co n trov ersia l as a n y th in g in lin g u istics n o w a d a y s. U n l e s s , it is eq u iv a len t to w h a t I h ave ca ",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Conclusions",
                "sec_num": "4"
            },
            {
                "text": "6Proceedings of NODALIDA 1989, pages 86-100",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            },
            {
                "text": "Proceedings of NODALIDA 1989",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            }
        ],
        "back_matter": [],
        "bib_entries": {
            "BIBREF1": {
                "ref_id": "b1",
                "title": "A Two-level Morphological Analysis of Japanese",
                "authors": [
                    {
                        "first": "Alain",
                        "middle": [],
                        "last": "",
                        "suffix": ""
                    },
                    {
                        "first": "Yukio",
                        "middle": [],
                        "last": "Sasaki",
                        "suffix": ""
                    }
                ],
                "year": 1983,
                "venue": "Texas Linguistic Forum",
                "volume": "22",
                "issue": "",
                "pages": "229--252",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Alain, Yukio Sasaki. 1983. A Two-level Morphological Analysis of Japanese. Texas Linguistic Forum, 22:229-252.",
                "links": null
            },
            "BIBREF2": {
                "ref_id": "b2",
                "title": "Morphology with Two-Level Rules and Negative Rule Features",
                "authors": [
                    {
                        "first": "John",
                        "middle": [],
                        "last": "Bear",
                        "suffix": ""
                    }
                ],
                "year": 1988,
                "venue": "Pro ceedings o f the 12th International Conference on Computational Linguistics",
                "volume": "",
                "issue": "",
                "pages": "2--31",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Bear, John. 1988. Morphology with Two-Level Rules and Negative Rule Features. Pro ceedings o f the 12th International Conference on Computational Linguistics:2&- 31, COLING, Budapest.",
                "links": null
            },
            "BIBREF3": {
                "ref_id": "b3",
                "title": "Paradigmatic Morphology. Proceedings o f the Forth Conference o f the European Chapter",
                "authors": [
                    {
                        "first": "Jonathan",
                        "middle": [],
                        "last": "Calder",
                        "suffix": ""
                    }
                ],
                "year": 1989,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "58--65",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Calder, Jonathan. 1989. Paradigmatic Morphology. Proceedings o f the Forth Conference o f the European Chapter o f the Association for Computational Linguistics:58-65, ACL, Manchester.",
                "links": null
            },
            "BIBREF4": {
                "ref_id": "b4",
                "title": "DKIMMO: A Morphologiciil Analyzer using Two-Level Rules",
                "authors": [
                    {
                        "first": "Mary",
                        "middle": [],
                        "last": "Dalrymple",
                        "suffix": ""
                    },
                    {
                        "first": "Lauri",
                        "middle": [],
                        "last": "Karttunen",
                        "suffix": ""
                    },
                    {
                        "first": "Sami",
                        "middle": [],
                        "last": "Shaio",
                        "suffix": ""
                    }
                ],
                "year": 1987,
                "venue": "Computational Morphology, Course Script LI283",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Dalrymple, Mary, Lauri Karttunen and Sami Shaio. 1987. DKIMMO: A Morphologiciil Analyzer using Two-Level Rules. In R. Kaplan and L. Kcirttunen: Computational Morphology, Course Script LI283, 1987 Linguistic Institute, Stanford University, June 29-August 7, 1987.",
                "links": null
            },
            "BIBREF6": {
                "ref_id": "b6",
                "title": "Three seductions o f computational psycholinguistics",
                "authors": [
                    {
                        "first": "Ronald",
                        "middle": [],
                        "last": "Kaplan",
                        "suffix": ""
                    }
                ],
                "year": 1987,
                "venue": "Linguistic Theory and Computer Applications:149-l88",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Kaplan, Ronald. 1987. Three seductions o f computational psycholinguistics. In P. Whitelock et al. (Eds.), Linguistic Theory and Computer Applications:149-l88, Academic Press, New York.",
                "links": null
            },
            "BIBREF7": {
                "ref_id": "b7",
                "title": "TW O L: A Compiler for Two-Level Phonological Rules",
                "authors": [
                    {
                        "first": "Lauri",
                        "middle": [],
                        "last": "Karttunen",
                        "suffix": ""
                    },
                    {
                        "first": "Kimmo",
                        "middle": [],
                        "last": "Koskenniemi",
                        "suffix": ""
                    },
                    {
                        "first": "Ronald",
                        "middle": [],
                        "last": "Kaplan",
                        "suffix": ""
                    }
                ],
                "year": 1987,
                "venue": "Computational Morphology, Course Script LI283",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Karttunen, Lauri, Kimmo Koskenniemi and Ronald Kaplan. 1987. TW O L: A Compiler for Two-Level Phonological Rules. In R. Kaplan and L. Karttunen: Computational Morphology, Course Script LI283, 1987 Linguistic Institute, Stanford University, June 29-August 7, 1987.",
                "links": null
            },
            "BIBREF8": {
                "ref_id": "b8",
                "title": "A Two-Level Morphological Analysis of English",
                "authors": [
                    {
                        "first": "Lauri",
                        "middle": [],
                        "last": "Karttunen",
                        "suffix": ""
                    },
                    {
                        "first": "Kent",
                        "middle": [],
                        "last": "Wittenburg",
                        "suffix": ""
                    }
                ],
                "year": 1983,
                "venue": "Texas Linguistic Forum",
                "volume": "22",
                "issue": "",
                "pages": "217--228",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Karttunen, Lauri and Kent Wittenburg. 1983. A Two-Level Morphological Analysis of English. Texas Linguistic Forum, 22:217-228.",
                "links": null
            },
            "BIBREF9": {
                "ref_id": "b9",
                "title": "Finite-state description of Semitic mor phology: A case study of ancient Accadian",
                "authors": [
                    {
                        "first": "Laura",
                        "middle": [],
                        "last": "Kataja",
                        "suffix": ""
                    },
                    {
                        "first": "Kimmo",
                        "middle": [],
                        "last": "Koskenniemi",
                        "suffix": ""
                    }
                ],
                "year": 1988,
                "venue": "Proceedings o f the 12th International Conference on Computational Linguistics",
                "volume": "",
                "issue": "",
                "pages": "318--315",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Kataja, Laura and Kimmo Koskenniemi. 1988. Finite-state description of Semitic mor phology: A case study of ancient Accadian. Proceedings o f the 12th International Conference on Computational Linguistics:318-315, COLING, Budapest. Computational Linguistics -Reykjavik 1989",
                "links": null
            },
            "BIBREF10": {
                "ref_id": "b10",
                "title": "A Two-level Morphological Analysis of Rumanian",
                "authors": [
                    {
                        "first": "Robert",
                        "middle": [],
                        "last": "Khan",
                        "suffix": ""
                    }
                ],
                "year": 1983,
                "venue": "Texas Linguis tic Forum",
                "volume": "22",
                "issue": "",
                "pages": "253--270",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Khan, Robert. 1983. A Two-level Morphological Analysis of Rumanian. Texas Linguis tic Forum 22:253-270.",
                "links": null
            },
            "BIBREF11": {
                "ref_id": "b11",
                "title": "Two-level morphology: A general computational model for word-form recognition and production",
                "authors": [
                    {
                        "first": "Kimmo",
                        "middle": [],
                        "last": "Koskenniemi",
                        "suffix": ""
                    }
                ],
                "year": 1983,
                "venue": "Publication",
                "volume": "11",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Koskenniemi, Kimmo. 1983. Two-level morphology: A general computational model for word-form recognition and production. Publication 11, Department of General Linguistics, University of Helsinki, Helsinki.",
                "links": null
            },
            "BIBREF12": {
                "ref_id": "b12",
                "title": "A Two-level Description of Old Church Slavonic Morphology. Scando-Slavica",
                "authors": [
                    {
                        "first": "Jouko",
                        "middle": [],
                        "last": "Lindstedt",
                        "suffix": ""
                    }
                ],
                "year": 1984,
                "venue": "",
                "volume": "30",
                "issue": "",
                "pages": "165--189",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Lindstedt, Jouko. 1984. A Two-level Description of Old Church Slavonic Morphology. Scando-Slavica, 30:165-189.",
                "links": null
            },
            "BIBREF13": {
                "ref_id": "b13",
                "title": "Psychological reality in phonology: A theoretical study",
                "authors": [
                    {
                        "first": "Per",
                        "middle": [],
                        "last": "Lineu",
                        "suffix": ""
                    }
                ],
                "year": 1979,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "LineU, Per. 1979. Psychological reality in phonology: A theoretical study. Cambridge University Press, Cambridge.",
                "links": null
            },
            "BIBREF14": {
                "ref_id": "b14",
                "title": "Morphology",
                "authors": [
                    {
                        "first": "P",
                        "middle": [
                            "H"
                        ],
                        "last": "Matthews",
                        "suffix": ""
                    }
                ],
                "year": 1974,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Matthews, P. H. 1974. Morphology. Cambridge University Press, Cambridge.",
                "links": null
            },
            "BIBREF15": {
                "ref_id": "b15",
                "title": "Abstrakta fonem och modeller som f\u00f8rst\u00e5r bara \"modellska",
                "authors": [
                    {
                        "first": "Martti",
                        "middle": [],
                        "last": "Nyman",
                        "suffix": ""
                    }
                ],
                "year": 1988,
                "venue": "The Linguistic Association of Finland",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Nyman, Martti. 1988. Abstrakta fonem och modeller som f\u00f8rst\u00e5r bara \"modellska\" . The Linguistic Association of Finland, Helsinki.",
                "links": null
            },
            "BIBREF16": {
                "ref_id": "b16",
                "title": "Studies of Grammars, Formalisms, and Parsing. Rapporter fr\u00e5n Spr\u00e5kdata, 25",
                "authors": [
                    {
                        "first": "Anna",
                        "middle": [],
                        "last": "S\u00e5gvall-Hein",
                        "suffix": ""
                    },
                    {
                        "first": "Mats",
                        "middle": [],
                        "last": "Dahll\u00f6f",
                        "suffix": ""
                    },
                    {
                        "first": "Sofia",
                        "middle": [],
                        "last": "H\u00f6rmander",
                        "suffix": ""
                    }
                ],
                "year": 1989,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "S\u00e5gvall-Hein, Anna, Mats Dahll\u00f6f and Sofia H\u00f6rmander. 1989. Studies of Grammars, Formalisms, and Parsing. Rapporter fr\u00e5n Spr\u00e5kdata, 25. Department of Compu tational Linguistics, University of Gothenburg.",
                "links": null
            },
            "BIBREF17": {
                "ref_id": "b17",
                "title": "Separating Linguistic Analyses from Linguistic Theories",
                "authors": [
                    {
                        "first": "Stuart",
                        "middle": [],
                        "last": "Shieber",
                        "suffix": ""
                    }
                ],
                "year": 1987,
                "venue": "U. Reyle and C. Rohrer {Eds.)Natural Language Processing and Linguistic Theories, D.Reidel",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Shieber, Stuart. 1987. Separating Linguistic Analyses from Linguistic Theories. In U. Reyle and C. Rohrer {Eds.)Natural Language Processing and Linguistic Theories, D.Reidel, Holland.",
                "links": null
            },
            "BIBREF18": {
                "ref_id": "b18",
                "title": "Constituent-Based Morphological Pars ing: A New Approach to the Problem of Word-Recognition",
                "authors": [
                    {
                        "first": "Richard",
                        "middle": [],
                        "last": "Sproat",
                        "suffix": ""
                    },
                    {
                        "first": "Barbara",
                        "middle": [],
                        "last": "Brunson",
                        "suffix": ""
                    }
                ],
                "year": 1987,
                "venue": "Proceedings o f the 25th Annual Meeting o f the Association fo r Computational Linguistics65-72, ACL",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Sproat, Richard and Barbara Brunson. 1987. Constituent-Based Morphological Pars ing: A New Approach to the Problem of Word-Recognition. Proceedings o f the 25th Annual Meeting o f the Association fo r Computational Linguistics65-72, ACL, Stanford, CA.",
                "links": null
            },
            "BIBREF19": {
                "ref_id": "b19",
                "title": "Bulgarian Nominal Inflection in the Two-Level Framework. Manu script",
                "authors": [
                    {
                        "first": "Jordan",
                        "middle": [],
                        "last": "Zlatev",
                        "suffix": ""
                    }
                ],
                "year": 1988,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Zlatev, Jordan. 1988. Bulgarian Nominal Inflection in the Two-Level Framework. Manu script, Department o f Linguistics, University of Stockholm.",
                "links": null
            }
        },
        "ref_entries": {
            "FIGREF0": {
                "uris": null,
                "num": null,
                "text": "n d sim p licity, o p in io n s are b e in g e x p re sse d th a t a m o r p h o lo g ic a l m o d e l sh o u ld b e gen eral, (u n d e r s to o d as u n iv e rs a l), e x p la n a to r y a n d e v en p sy ch o lo g ica lly real. N o w fa r fr o m d o u b tin g th e p la u s ib ility o f th e se re q u irem en ts, I feel th a t th e y ten d t o p la ce th e m o d e ls o f h u m a n la n g u a g e p r o v id e d b y th e field o f c o m p u ta tio n a l lin gu istics in a rath er u n fa v o u ra b le lig h t. T h is is e s p e c ia lly relevan t fo r c o m p u ta tio n a l m o rp h o lo g y , w h ich o n ly d u rin g th is d e c a d e seem s t o h a v e \" ste p p e d o u t fr o m th e c ra d le \" , as fo r e x a m p le L a rs B o r in ( p .c .) has im p lie d . A n d inste^ui o f b e in g b lin d ly critica l a n d s c e p tic a l to w a rd s its p o te n tia ls , (w h ich \" lin gu ists p r o p e r \" o fte n ten d t o b e tow a rd s c o m p u ta tio n a l lin g u istics in g en era l, as a fo r m o f s e lf-d e fe n c e ), is n ' t it b est t o w a tch its first step s ca refu lly , w ith a h e lp fu l h a n d w h ere it ca n b e le n t? I f in th e p re v io u s p a ra g ra p h I h a v e su g g e ste d th e p ic tu r e o f c o m p u ta tio n a l m o rp h o lo g y , a n d m o re c o n c r e te ly o f it 's b e s t k n o w n re p re se n ta tiv e , th e tw olev el m od el, (first p rese n te d in (K o s k e n n ie m i 1 983) a n d m o s t o fte n a b b r e v ia te d T W O L ), as a clu m sy , stu m b lin g b a b y , th e n I h a v e g o n e t o o fa r in m y m a n n er o f ex p re ssio n . N o th in g ca n b e fu rth e r fr o m th e tru th c o n s id e r in g th e e n o r m o u s a m o u n t o f a tte n tio n a n d su b seq u en t w o rk th a t K o s k e n n ie m i's d isse rta tio n u n lea sh ed . H axdly a c o n fe re n ce ca n g o b y -in clu d in g th is o n e -w ith o u t a few c o n trib u tio n s p o in tin g o u t T W O L 's a ch ie v em en ts-o r d e ficien cie s, a n d in th e b e st ca se o ffe rin g im p ro v e m e n ts o r a lte rn a tiv es, e .g . (S p r o a t a n d B r u n s o n 1987, B e a r",
                "type_str": "figure"
            },
            "FIGREF1": {
                "uris": null,
                "num": null,
                "text": "d o u b t fall a t th e h a n d o f th e th e o r e tic a l lin gu ist w h o w ill n o t fa il t o see th e in a d eq u a cy o f B e a r 's r e in tr o d u c in g th e n o tio n o f \" n e g a tiv e ru le fe a tu re s\" , o r o f C a ld e r 's \" strin g e q u a tio n s \" . A s th e la st a u th o r h im s e lf c a re fu lly sta te s: \" . . . o n e m a y ju s tifia b ly h a v e re serv a tion s a b o u t in tr o d u c in g strin g e q u a tio n s in to lin g -' -^tic d e s c r ip tio n s .\" (C a ld e r 1 9 8 9 :6 2 ). In th is p a p e r I w ish t o p r o p o s e w h a t I th in k is a m o r e \" c o n s tr u c tiv e \" v ie w o f th e aim s o f c o m p u ta tio n a l m o r p h o lo g y , w h ich is a ls o m o re o r less a p p lic a b le to th e field o f (c o m p u ta t io n a l) n a tu ra l la n g u a g e p r o c e s s in g in g en era l. I w ill argu e th a t th ere are a n u m b e r o f p r o p e r tie s , w h ich ca n h elp us c o m p a r e , ev a lu a te an d d e v e lo p m o d e ls in a m o r e s h o r t-te r m p e r s p e c tiv e s o th a t o n e n ee d n o t n ecessa rily b e o v e rw h e lm e d b y th e \" th e o r e tic a l a rg u m e n t\" fr o m th e b e g in n in g . In m y o p in io n a c o m p u ta tio n a l m o d e l th a t fin d s th e b e s t c o m b in a tio n o f th ese p ro",
                "type_str": "figure"
            },
            "FIGREF2": {
                "uris": null,
                "num": null,
                "text": "I w o u ld like t o co n tin u e o n th is lin e o f th o u g h t w ith o n e su b sta n tia l d ifferen ce: w h ile S h ieb er d iscu sses m o d e ls in th eir p r o p e r ty o f b e in g \" c o m p u te r t o o ls fo r lin gu istics\" , I reg a rd th em as p o te n tia l ca n d id a te s fo r b e c o m in g la n g u a g e th e orie s o n th eir o w n . T h is d ifferen ce is illu stra te d in th e c h o ic e o f m o d e l t o e x e m p lify th e issues u n d er d iscu ssio n : in S h ie b e r' s ca se th is is th e fo rm a lis m o f P A T R -I I , w h ile I w ill use th e tw o -le v e l m o d e l. I w ill b e p r e s u p p o s in g a t lea st s o m e p r e v io u s k n o w le d g e o f it.",
                "type_str": "figure"
            },
            "FIGREF3": {
                "uris": null,
                "num": null,
                "text": "C o m p u t a tio n a l lin g u istics (p a r tly b e c a u s e o f u tilita ria n rea son s) is m u ch m ore in clin e d t o u se th e tr ia l-a n d -e r r o r a p p r o a c h , sta rtin g w ith a fra g m en t an d then a u g m e n tin g it; ta k in g s o m e ca te g o r ie s fo r g ra n ted , (p h o n e m e s , fo r e x a m p le ) as \" w o r k in g h y p o th e s e s \" , i f th e y fa c ilita te th e ov era ll w ork o f th e sy stem . T h is is so b e c a u s e th e s h o r t-te r m g o a l o f c o m p u ta tio n a l lin gu istics is th e c o n s tr u c tio n o f a n a tu ra l lan gu age p ro c e s s in g s y s te m , n o m a tte r i f it d o e s o r d o e s n o t m o d e l hum an",
                "type_str": "figure"
            },
            "FIGREF4": {
                "uris": null,
                "num": null,
                "text": "d th e o re tica l m o d e ls sh o u ld n o t b e co n s id e r e d o n a par. A co m p u ta tio n a l m o d e l is b o th less and m o r e th a n a th e o r e tic a l o n e . L ess, b e c a u s e it is th e b a c k b o n e o f a sy stem and th u s is s u b je c te d t o th e lim ita tio n s I m e n tio n e d a b o v e , i.e. w o rk in g h y p oth eses, fra g m en ts e tc . M o r e , b e c a u s e i f it is fle x ib le en o u g h it c o u ld p e rm it s e v e r a l th e o r e tic a l m o d e ls t o b e im p le m e n te d (sim u la te d ) w ith in it. S o th e qu estion w h e th e r T W O L is a m o r p h o lo g ic a l m o d e l o r n o t, is n ot all th a t relevan t. A s to",
                "type_str": "figure"
            },
            "FIGREF5": {
                "uris": null,
                "num": null,
                "text": "I f a t lea st th e sh o rt te rm a im s o f c o m p u ta tio n a l a n d th e o re tica l lin gu istics s p lit, th en w h a t a re t o b e th e c rite ria fo r , let us say \" e v a lu a tin g \" , c o m p u ta tio n a l m o d e ls (th e o r ie s , fo rm a lism s-th e te r m in o lo g y va ries) fo r m o r p h o lo g y in p a rticu la r, a n d n a tu ra l la n g u a g e in g e n era l? T h e r e is n o sim p le an sw er t o th is q u e s tio n . A s a h a lf y e a r 's su rv e y o f th e relevan t litera tu re , r e p o r te d in (Z la te v et al. 1 9 8 9 ) a n d (S a g v a ll-H ein et al. 1 9 8 9 ), has m a n a g ed t o c o n v in c e us-o p in io n s d iffer. W e c a m e t o b e lie v e th a t in o r d e r t o c o m e t o m o r e a b s tr a c t th in g s such as d e s id e r a ta , re q u irem en ts e tc . fo r th e m o d e ls , o n e sh o u ld sta rt w ith s om eth in g m o r e c o n c r e te . T h e k ey lies in w h a t I m e n tio n e d a b o v e w as o n e o f th e first aim s o f c o m p u ta tio n a l lin g u istics, a n d d e fin ite ly th e first o f its m o re p r a c tic a lly o ri e n te d s u b -b r a n c h . N a tu ra l L a n g u a g e P r o c e s s in g (N L P ): th e cre a tio n o f an N L P s y stem .",
                "type_str": "figure"
            },
            "FIGREF6": {
                "uris": null,
                "num": null,
                "text": "an d (2 ) to g e th e r m ak e up th e c o m p u ta tio n a l m o d e l o r -u sin g a te rm m o re n eu tral t o th e c o m p u ta tio n a l/t h e o r e t ic a l d ic h o to m y w h ich I m y s e lf in tr o d u c e d -an N L P fra m ew o rk . (3 ) is th e lan gu age d e scrip tio n . T h e th ree are as I sa id in te rd ep en d en t, b u t t o differen t d e g rees in d ifferen t sy stem s.",
                "type_str": "figure"
            },
            "FIGREF7": {
                "uris": null,
                "num": null,
                "text": "m a in a d v a n ta g e in v ie w in g c o m p u ta tio n a l m o d e ls o f n a tu ra l la n g u a g e a n d o f m o r p h o lo g y in p a r tic u la r as N L P fra m ew o rk s c o m e s fr o m th e fa c t th a t it is p o ssib le t o fo rm u la te r e la tiv e ly cle a rly w h a t g o a ls N L P sy ste m s s h o u ld a im at. T h e n o n e c o u ld c o n tin u e \" b o t t o m -u p \" t o s ta te \" c rite ria \" o n h o w th e m o d e ls sh o u ld b e s h a p e d in o r d e r t o c o r r e s p o n d t o th ese g o a ls. C o n s e q u e n tly th ese are crite ria o f a p r a c tic a l n a tu re w h ich a re n o t \" th e o r e tic a lly b o u n d \" t o b e g in w ith . M o s t in terestin g ly , h o w e v e r, th e y h ave im p lic a tio n s w h ich are h ig h ly c o m p a tib le w ith lin g u istica lly m o tiv a te d c o n s id e ra tio n s . I w ill c o m e t o th is in th e last se c tio n . A c o m p u ta tio n a l m o d e l su ch as T W O L m a y b e seen as p r o v id in g th e fra m e w o rk fo r an N L P sy stem . It still rem a in s t o b e \" filled \" w ith th e c o n c r e te lin g u istic k n o w led g e . N o w th e first q u e stio n th a t arises is: h o w m u ch k n o w le d g e ca n b e ex p ressed in th e fra m e w o rk ? T h e first g o a l fo r an N L P s y ste m is th a t th is k n o w l ed g e is su fficien t fo r cu rren t p u rp o s e s , o r a lte rn a tiv e ly fo r m u la te d , th a t it has su fficien t cov era g e.",
                "type_str": "figure"
            },
            "FIGREF8": {
                "uris": null,
                "num": null,
                "text": "d e scrip tio n o f a certa in fra g m en t o f o n e o r several la n g u a g es ca n b e m a d e so th a t th e sy ste m \" w o rk s\" as in te n d e d w ith r e sp e ct t o th is fra g m e n t, th e n th e fraonew ork ca n b e re g a rd e d as ex p re ssiv e e n o u g h in re la tio n t o th is fra g m e n t. T h u s o n e m a y say th a t an N L P fra m e w o rk is w eakly co m p le te (in S h ie b e r 's te r m in o lo g y ) if a n d o n ly i f it p ro v id e s a sy ste m w ith th e lin g u istic co v e r a g e n ecessa ry fo r th e g iv en p u rp o se s. T W O L has b een a p p lie d t o su b sta n tia l fra g m e n ts o f th e in fle ctio n a l m o r p h o l",
                "type_str": "figure"
            },
            "FIGREF9": {
                "uris": null,
                "num": null,
                "text": "n d d e riv a tion a l (w h e re th is d is tin c tio n e x is ts ), i.e. th a t it is a u n iv ersa l m o r p h o lo g ica l m o d e l. It is ra th er a m a tte r o f d eg ree: T W O L is \" b e tt e r \" th a n m o s t o th e r m o d e ls b e ca u se it has b e e n a p p lie d t o la rg er fra g m e n ts o f sin g le la n g u a g es, e.g . \" an (a lm o s t) fu ll d e s c r ip tio n ( o f all th e fo rm s o f all in fle ctio n a l t y p e s )\" (K o s k e nniem i 1 9 8 3 :1 2 5 ) a n d b e c a u s e it has b een a p p lie d to m o r e la n g u a g es. B u t th en , w h a t m o re is n e e d e d ? T h e fa c t th a t th e m o r p h o lo g y o f fo r e x a m p le K u b a c h i (cf. Joh a n n essen , th is v o lu m e ) y ie ld s difficu lties, d o e s n ' t m a k e T W O L a less s u ita b le fra m ew o rk fo r th e d e s c r ip tio n o f, le t 's say, B u lg a ria n in fle ctio n . T h is o n ly m ea n s th a t th e m o r p h o lo g ie s o f th e tw o la n g u a g es are d iffe re n t-th e o p p o s it e w o u ld b e su rp risin g. H o w ev er, i f o n e b y a \" g en era l\" fra m e w o rk m ea n s o n e th a t ca n p r o v id e a d e q u a te d e s crip tio n s fo r all la n g u a g e ty p e s: a g g lu tin a tin g , is o la tin g , in fle ctin g , e tc. th en m o re is t o b e d esired . T h is falls, in m y o p in io n , n o t u n d e r th e g o a l o f co v e ra g e b u t o f flex ib ility , w h ich w ill b e d iscu ssed fu rth e r o n . L e t us b e m o re c o n c r e te . In (Z la t e v 1 9 8 8 ) I h a v e g iv en w h a t I th in k is a co m p le te d e scrip tio n o f B u lg a ria n n o m in a l in fle ctio n in term s o f th e o rig in a l T W O L , i.e. as p re se n te d in (K o s k e n n ie m i 1 9 8 3 ). B u lg a ria n m o r p h o lo g y is v ery w ell d e v e lo p e d a n d p o s e s s o m e n o n -tr iv ia l p ro b le m s fo r a n y lin gu istic d e scrip tio n , c o m p u ta tio n a l o r n o t, su ch as e x te n s iv e a llo m o r p h y a n d m o r p h o p h o n e m ic a lte rn a tio n s w ith in th e ste m s. T W O L h as p r o v e d q u ite s a tis fa c to r y in d e scrib in g b o t h , w ith its fin ite -s ta te le x ic o n a n d tw o -le v e l rules, resp e ctiv e ly . T h e d e m o n s tr a tiv e p r o n o u n s , h o w e v e r, d is p a y an \" irre g u la r\" in tern a l in fle ctio n , w h ich in th e o r ig in a l (P a sced ) fo r m a t o f th e le x ic o n g iv e s n o o th e r o p p o r t u n it y fo r d e scrip tio n th a n th e fo llo w in g , w h ich is fa r & o m e le g a n t. o-a-ov-e/P \"PRON DEM IDENT\" w ith ' t ' as th e \" in varian t s te m \" (I h a v e stick e d to th e p rin cip le : \" O n e en try p er S te m \" s o as t o a v o id m a sk in g s o m e p r o b le m a tic a l areas th ro u g h listin g ) andth e c o n tin u a tio n cla ss o -a -o v -e / P w h ich is th e n a m e o f a m in i-le x ic o n w ith th e fo llo w in g c o n te n t:",
                "type_str": "figure"
            },
            "FIGREF10": {
                "uris": null,
                "num": null,
                "text": "f th is h a d b e e n th e reg u la r p a tte r n fo r in fle ctio n in B u lg a ria n , th en a p o ssib le c o m p u ta tio n a l d e s c r ip tio n in th e fo r m o f a sy ste m o f in terse ctin g le x ic o n s -as th o s e p re se n te d in ( K a t a ja a n d K o s k e n n ie m i 1988) fo r th e n o n -c o n c a te n a tiv e m o r p h o lo g y o f S e m itic la n g u a g es-w o u ld h ave b een n ecessa ry (a n d p r o b a b ly s u fficie n t). H o w e v e r, sin ce th e n u m b e r o f m in i-le x ico n s o f th e k in d sh ow n a b o v e is 5 a lto g e th e r a n d a ll o th e r ty p e s fa ll n e a tly in to th e fin ite -s ta te p a tte rn , a c o m p r o m is e seem s t o b e th e b e s t s o lu tio n : I co n sid e r T W O L ex p re ssiv e en ou g h , i.e. su fficien t fo r cu rren t p u r p o s e s , a n d d e c id e t o lea v e th e d e s c r ip tio n a t th a t. W h a t i f I d e c id e t o tre a t d e riv a tio n a l m o r p h o lo g y as w ell? F iv e classes o f B u lg a ria n p r o -fo r m s seem t o b e re a d ily d e s c r ib a b le as a d e riv a tio n a l p a ttern w h ich is s o m e th in g o f th e so rt: t is, th e in te r r o g a tiv e p r o -fo r m s ( B ) , a ct as th e \" b a se\" , w h ich to g e th er w ith th e a p p r o p r ia te \" p r e fix \" , b u ild r e s p e c tiv e ly in d efin ite, n eg a tiv e a n d gen eral iz in g p r o -fo r m s ( A ) , a n d w ith th e \" su ffix\" ' t o ' (w h ich is a ctu a lly th e p o s tp o n e d d e fin ite a rticle fo r n o u n s a n d a d je c tiv e s o f n eu te r g e n d e r )-rela tiv e p r o f o r m s ( C ) . H ow ev er, i f w e tr y t o ex p ress th is s im p le p a tte rn in a fin ite -s ta te le x ico n th e n w e w ill a ls o d e r iv e u n g ra m m a tica l w o r d -fo r m s su ch as *n A ka kto, *nikakyvto e t c .",
                "type_str": "figure"
            },
            "FIGREF11": {
                "uris": null,
                "num": null,
                "text": "N o w d o e s th is m ea n th a t w e h a v e fo u n d a p o in t w h e re T W O L is n o t sufiBcient in term s o f co v e r a g e a n d a n a rg u m en t th a t it is in a p p lic a b le t o B u lg a ria n as w ell as p o s s ib ly th e d e riv a tio n a l m o r p h o lo g y o f m o s t la n g u a g es? T o s o m e e x te n t-y es. F o r p ra c tic a l p u rp o s e s w e m a y d o u b le th e en tries o f ty p e B in th e le x ic o n , s o th a t w e h a v e B ' a n d th e n c o n n e c t th e m in i-le x ic o n s , t o g et A B a n d B ' C (fo r e x a m p le ). B u t th is is a k in d o f \" s o lu tio n \" th a t w o u ld lerid us b a ck t o w h e re w e s ta rte d , a n d it is in s o m e sen se e v en w o rse th a n listin g th e differen t w o r d -fo r m s -it is a b su rd th a t w e s h o u ld h a v e t o g o a ll th is w a y o n ly t o sta rt d u p lic a tin g en tries. (H ere I 'm n o t c o n c e r n e d w ith m a tte rs o f e fficie n cy -b u t th ese a re o f c o u rs e m o re th an relevan t as w ell.) T h e r e a re, h ow ever, tw o o th e r m u ch b e tt e r w a y s o u t. O n e w o u ld b e t o re p la ce th e fin ite -sta te le x ico n c o m p o n e n t w ith a p h ra s e -s tru ctu re o n e , w h ich fu rth e r m o re ca n u se a fe a tu re -m a tch in g (u n ific a tio n ) m ech a n ism w h ich w o u ld g u a ra n te e th a t o n ly th e g ra m m a tica l fo rm s are g e n e ra te d . F o r e x a m p le th e p r o b le m I m en tio n e d a b o v e ca n b e resolv ed th e fo llo w in g w ay:(1) PRO(IND) -> n\u00c5 + PRO(INT) PRO(NEG) -> ni + PRO(INT) PRO(GEN) -> vsA + PRO(INT) PRO(REL) -> PRO(INT) + to A n a ltern a tiv e-w ith o u t in cre a sin g th e e x p ressiv e p o w e r o f th e fo r m a lis mis to use th e m o d e l's tw o -le v e l ru les in o r d e r t o b lo c k o u t u n g ra m m a tica litie s . In th e ca se a b o v e o n e m u st use a t lea st tw o \" d ia c r itic ch a ra cte rs \" , le t us say, Q a n d # (w h ich m u st b e c le a rly d efin ed as b e a re rs o f m o r p h o lo g ic a l fea tu res an d h ave n o th in g t o d o w ith p h o n o lo g y ) a n d a s s o c ia te th e m w ith th e en tries o f typ e A a n d C , resp ectiv ely . T h e n a ru le c a n b e s ta te d w h ich w o u ld p rev e n t th eir c o -o c c u r r e n c e , (th e o p e r a to r / < -m ea n s \" is d isa llo w e d \" a n d w h a t fo llo w s",
                "type_str": "figure"
            },
            "FIGREF12": {
                "uris": null,
                "num": null,
                "text": "e th e sa m e e ffe ct, a n d w h ich o n e w o u ld b e p re fe rred is la rg ely a m a tte r o f h o w th e y in flu en ce th e g o a ls t o b e d iscu ssed b e lo w , n a m e ly efficien cy a n d a u g m en ta b ility .",
                "type_str": "figure"
            },
            "FIGREF13": {
                "uris": null,
                "num": null,
                "text": "E fficie n cy is s o m e th in g th a t c o n c e r n s n o t o n ly N L P sy ste m s fo r p r a c tic a l p u r p o s es, b u t th e o re tica l o n e s as w ell, sin ce all in terestin g a p p lic a tio n s o f c o m p u ta tio n a l tech n iq u es t o n a tu ra l la n g u a g es in v o lv e fra g m e n ts th a t g o b e y o n d v o c a b u la r ie s o f several h u n d red w o rd s a n d a p re d e te rm in e d n u m b e r o f sen ten ces.It is n o t h a rd t o b e lie v e th a t it is ju s t th is c r ite rio n th a t has been th e m ain rea son fo r T W O L ' s p o p u la r it y ra th er th a n its lin g u istic ch a ra cteristics.T h e re s t r ic tiv e n e s s o f th e fo r m a lis m g iv e s th e o p p o r tu n ity o f e x tr e m e ly efficien t im p le m e n ta tio n s in w h ich th e le x ic o n h as th e fo r m o f a w o r d tree a n d th e ru lesfin ite -s ta te tra n sd u ce rs. T h is h as b r o u g h t a b o u t th e p o s s ib ility o f c o n s tru ctin g s y ste m s w ith le x ic o n s o f ten s o f th o u s a n d s o f stem s w h ich ca n p ro ce s s te x t co r p o r a a n d retu rn a n a ly ses w ith m o r p h o lo g ic a l fea tu res a t a sp en d o f u p t o 100 w o r d -fo r m s a m in u te (F re d K a rlsso n p .c .) I d o n ' t in te n d t o in d u lg e in th is m a tte r sin ce I 'm n o e x p e r t. S till, th e in terde p e n d e n c y o f s y s te m e fficie n cy a n d o th e r g o a ls m ust b e p o in te d o u t. F or exrim ple, w h e n it c o m e s t o c h o o s in g b etw een th e tw o solu tion s t o th e co v e ra g e p ro b le m w h ich I d iscu sse d a b o v e , o n e w o u ld p r o b a b ly a d o p t th e s e c o n d a ltern a tiv e-th a t w ith th e d isa llo w in g rule-i f a b a n d o n in g th e fin ite-sta te fo r m a t o f th e le x ico n is lik ely t o s lo w d o w n im p le m e n ta tio n s d ra stica lly . A n d th is w o u ld b e a rea son a b le m o v e -as lo n g as it d o e s n ' t g e t in th e w a y o f th e n e x t g o a l.",
                "type_str": "figure"
            },
            "FIGREF14": {
                "uris": null,
                "num": null,
                "text": "A n N L P s y ste m is sa id t o b e a u g m e n ta b le i f it ca n b e im p ro v e d w ith regard t o e a ch o f its s u b c o m p o n e n t s -fo rm a lism , p ro ce ssin g m ech a n ism a n d lin gu istic d e s c r ip tio n . E v en i f th e first tw o a re far fr o m u n ch a n g ea b le, th e y are n ev e rth e less m o r e sta b le th a n th e th ird , th e d e v e lo p m e n t o f w h ich is b y its n a tu re an in cre m e n ta l a n d inter^lctive p ro ce s s , w h ich g o e s th ro u g h lo o p s , d e a d en d s, p a r tia l s o lu tio n s , g r a d u a l g e n e ra liza tio n s e tc ., u n til it reach es a p ro v is io n a lly sta b le lev el, a jid th e n a g a in m u st b e su ch th a t it is p o s s ib le to d e v e lo p it w h en th e need arises. F o r th is re a son it is cen tra l th a t th e form a lism is p er s p ic u o u s -a q u ite in fo r m a l c r ite r io n , b u t n ev erth eless an im p o r ta n t on e . I h a v e a lre a d y d iscu sse d q u e stio n s p e rta in in g t o th e T W O L le x ic o n c o m p o n en t s o I w ill ta k e a fe w e x a m p le s fr o m th e \" h e a rt\" o f th e m o d e l-th e tw o-lev el ru les. B e fo r e a c o m p ile r fo r th em e x iste d , it w as a c u m b e r s o m e affair t o tran s la te in to tra n sd u ce rs e v en th e sim p lest rules. T h e T W O L c o m p ile r (K a rttu n e n e t al. 1 987) w as a g re a t a d v a n ce in th is r e s p e ct. It g iv es th e o p p o r tu n ity fo r a q u ite g en era l m o r p h o p h o n e m ic ru le t o b e fo rm u la te d as sim p ly as th is, (F r o n tV is d efin ed as I , E . ) :",
                "type_str": "figure"
            },
            "FIGREF15": {
                "uris": null,
                "num": null,
                "text": "le v e l)\" . It w ill a c c o u n t fo r e x a m p le fo r th e fo llo w in g p a irs. (T h e s e co n d e x a m p le sh ow s h o w p r o d u c t iv e th e ru le is!) L e x ica l rep resen ta tion : r y k a E (h a n d + PL) S u rfa ce rep resen ta tion : r y e e L e x ica l re p re se n ta tio n : h o t d o g I ( h o t d o g + PL) S u rfa ce rep resen ta tion : h o t d o z i O n e o f th e rea son s w h y su ch rules axe m o r e p e rs p icu o u s th a n fo r e x a m p le rules o f g e n era tiv e p h o n o lo g y is th a t th e y a re p u re ly d ecla ra tive s ta te m e n ts w h ich n eed n o t tak e in to co n s id e r a tio n a n y req u irem en ts o f o r d e r in g a n d th e c o m p le x in tera ction s th a t g o w ith it. It is first a fte r c o m p ila tio n (in t o fin ite -s ta te tra n s d u cers) th a t th e y g a in th e ir p r o c e d u r a l in te rp re ta tio n . A n o t h e r fa c t o r th a t m ak es it easier fo r a T W O L sy ste m t o b e a u g m e n te d is th e le x ic o n /r u le s d is tin c tio n itself, w h ich is a n e x a m p le o f th e p o s it iv e c o n seq u en ces o f m od u larity. F o r e x a m p le ev en i f n ew in fle ctio n a l a n d d e riv a tio n a l ty p e s are e v en tu a lly d isco v e re d , th e le x ic o n c a n b e ch em ged, b u t i f a ru le-as th e o n e a b o v e -is gen era l e n o u g h , th e n it n eed n o t b e \" tin k ered w ith \" a t all, b u t ca n sa fely a p p ly o n th e n ew le x ica l in fo r m a tio n . N o w , th e ru le as I s ta te d it a b o v e is a c tu a lly differen t fr o m th a t in (Z la t e v 1988) in th a t it d o e s n o t u se a n y d ia c r itic ch a ra cte rs w h ich h ave th e fu n c tio n o f \" m o rp h e m e b o u n d a r ie s \" , \" trig g ers\" , \" b lo c k e r s \" e tc . T h e p o in t is th a t su ch ch a ra cters w ork ag a in st th e p e r s p ic u ity o f th e fo rm a lis m a n d th u s a g a in st th e a u g m e n ta b ility o f th e sy ste m . F o r e x a m p le in (Z la t e v 1 9 8 8 :3 3 ) I w r o te : \" T h e o p e ra tio n a l le x ic o n ca n b e a u g m e n te d w ith n ew le x ica l stem s w h ich o n ly h ave to b e g iv en th e a p p r o p r ia te c o n tin u a tio n classes (a n d i f n e ce ssa ry t o u se th e d ia critics in th e righ t p o s it io n s )\" . It is ju s t th e se \" rig h t p o s it io n s \" , w h ich w o u ld m ak e it s o h a rd fo r a n y o n e else th a n m y s e lf t o d e v e lo p th e sy stem .",
                "type_str": "figure"
            },
            "FIGREF16": {
                "uris": null,
                "num": null,
                "text": "h e g o a l o f fle x ib ility is c lo s e t o th a t o f a u g m e n ta b ility d iscu sse d a b o v e , b u t c o n ce rn s th e a b ility o f ch a n g e n o t o n ly fo r th e sake o f im p ro v e m e n t, b u t as a value in itself. T h e r e \u00a3ire d ifferen t levels o f flex ib ility . O n e is th e h a rd w a re d im e n sio n : a sy stem sh o u ld p re fe ra b ly b e in d e p e n d e n t o f a n y p a rticu la r ty p e o f m a ch in e. T h e recen t \"e m a n c ip a tio n \" o f th e T W O L c o m p ile r fr o m th e d e m a n d in g e n v iro n m e n t o f L isp -m a ch in es, i.e. th e e x is te n ce o f a c o m p ile r ru n n in g o n th e m o r e p o w e rfu l m od els o f A p p le M^lcintosh (K im m o K o s k e n n ie m i p .c .) m a y b e co n s id e r e d in this resp ect as a ste p in th e righ t d ir e c tio n . A n o th e r a sp e ct o f fle x ib ility is w ith reg a rd t o softw a re: a sy ste m sh o u ld n o t in a n y m a jo r d eg ree d e p e n d on a p a rticu la r p r o g r a m m in g la n g u a g e . T h e fa c t th a t m o d e rn p r o g r a m m in g la n g u a g es h a v e eq u iv a len t a b s o lu te e x p re ssiv e p o w e r,",
                "type_str": "figure"
            },
            "FIGREF17": {
                "uris": null,
                "num": null,
                "text": "n o ta tio n a lly s o (c f. S h ieb er 1 987) a n d it is s o m e tim e s e a sy t o fa ll fo r th e \" p r o ce d u ra l se d u c tio n \" o f c o m p u ta tio n a l lin g u istics th a t K a p la n (1 9 8 7 ) d iscu sses, e .g .t o d e p e n d o n P R O L O G 'S b a ck tra ck in g m e ch a n ism o r o n L is p 's ev a lu a tio n p ro ce d u re s. T W O L h as b e e n im p le m e n te d in P a s c a l (K o s k e n n ie m i 1 9 8 3 ), C o m m on L isp (G a je k et a l 1 9 8 3 ), In terlisp -D (D a lr y m p le et al 1 9 8 7 ) a n d C (K im m o K o s k e n n ie m i p .c .) w h ich is a sign th a t th e m o d e l is m o r e o r less in d ep e n d en t o f p r o g r a m m in g e n v iro n m e n t. B o th th ese a s p e cts o f fle x ib ility ca ll fo r tre a tin g hard-and so ftw a re in d ep en d en ce as a c rite rio n in itself. M a tte r s o f im p le m e n ta tio n , h ow ev er, are n o t o f p rim a ry in terest fo r us here. I w o u ld a c tu a lly w a n t t o \" s tre tch \" th e c o n c e p t o f fle x ib ility o f an N L P sy stem a n d in te rp ret is a s lin g u istic flex ib ility , o r in o th e r w o r d s : th e p ro p e r ty o f al lo w in g d ifferen t sty les o f d e s c r ip tio n . It is h ere th a t th e relev a n ce o f th e oretica l co n s id e r a tio n s is g re a te st. L e t us lo o k a g a in at T W O L in a little m o re detail. A s fa m ilia r, T W O L cein p r o v id e m o r p h o lo g ic a l d e s crip tio n s in th e sty le o f b o t h o f th e tr a d itio n a l m o r p h o lo g ic a l m o d e ls \" Item a n d A rra n g e m e n t\" ( l A ) and \" Ite m a n d P r o c e s s \" (I P ). F o r e x a m p le co n sid e rin g ru le (3 ) a n d th e ex a m p les in 3 .3 . o n e ca n d e s c r ib e th e sin g u la r a n d p lu ra l fo r m o f th e B u lg a ria n lexem e ' h o t d o g ' in th e fo llo w in g w a y (a ssu m in g th e fea tu re-v a lu e fo r m a t o f th e le x ico n p re se n te d as an o p tio n in (D a lr y m p le e t al 1 9 8 7 ), th o u g h w ith o u t th e fa cilita tin g d e v ic e o f \" te m p la te s \" ):",
                "type_str": "figure"
            },
            "FIGREF18": {
                "uris": null,
                "num": null,
                "text": "h ile (4 ) m a k es u se o n ly o f th e le x ic o n , (5 ) relies o n th e \" P a la ta liz a tio n o f V e la rs\" -ru le as w ell (p lu s a d e fa u lt m ech a n ism s ta tin g th a t n u m reciev es th e v a lu e sg, i f u n s p e c ifie d ).I f w e h a v e t o c o m p a r e th e tw o a lte rn a tiv e s, (5 ) seem s to b e b e tte r in al m o s t all r e sp e cts. It is n o t o n ly m o r e \" e le g a n t\" , it is sh o rte r, sim p ler an d as I a r g u e d in th e p r e v io u s s e c tio n th is t y p e o f d e s c r ip tio n is m o re p e rsp icu o u s and m o d u la r , th u s a s y s te m b a se d o n it w o u ld b e m o r e ea sily a u g m en te d . F rom a lin g u is tic p e r s p e c tiv e (4 ) w o u ld s im p ly d e s c r ib e ' h o t d o g /h o t d o z ' as a llo m o rp h s in co m p le m e n ta r y d is tr ib u tio n w h e re o n ly th e s e c o n d tak es p lu ra l w h ich is o b lig a to r y w h ile (5 ) w o u ld fu r th e r m o r e in c o r p o r a te th e p r o c e s s o f p a la ta liz a tio n in th e d e scrip tio n a n d th u s \" e x p la in \" th e a llo m o rp h y . T h is c o u ld p o s s ib ly im p ly th a t B u lg a ria n m o r p h o lo g y (a n d p r o b a b ly a n y m o r p h o lo g y w ith m o r p h o p h o n e m ic a lte rn a tio n s ) is m o r e re a d ily d e s c r ib a b le in term s o f IP th an l A . T h is is equ iv a len t t o sa y in g th a t it c o u ld b e a s s o c ia te d w ith a \" ty p o lo g ic a l p a r a m e te r \" w h ich d e te rm in e s th e m o s t a p p r o p r ia te sty le o f d e scrip tio n (c f. M a tth e w s 1 9 7 4 :1 6 3 ). H ow ev er, th e fa c t th a t th is is an o p tio n in T W O L , w h ich a lso p e r m its d e s c r ip tion s o f t y p e ( 4 ) -s u p p o s e d ly su fficien t fo r p u re ly a g g lu tin a tin g la n g u a g e s-is an o b v io u s a d v a n ta g e in term s o f flexib ility. S o T W O L c a n in pr^lctice fu lly \" m o d e l\" b o th I A a n d IP. T h e e x te n t t o w h ich th is is s o is so m e tim es o v e r lo o k e d b e c a u s e o f th e fa c t th a t all im p le m e n ta tio n s th a t I k n o w o f h ave used in p u t su ch as b o o k + s , in ste a d o f b o o k + PL d u rin g g e n e ra tio n , i.e. n eg lelecte d th e in fo r m a tio n in th e le x ic o n . T h is sh o u ld n o t b e co n sid e re d a d isa d v a n ta g e o f th e m o d e l itself, s in ce th e o n ly re a so n w h y it h a s n ' t b ee n im p lem e n ted is th a t u p t o n ow T W O L has n o t b e e n u sed fo r w o r d -fo r m p r o d u c tio n in a n y la rg er a p p lic a tio n . M o r p h o lo g ic a l c o n d itio n s a re fu rth e rm o re ex p ressib le (a n d are ex p ressed all th e tim e ) in th e c o n te x ts o f th e ru les th ro u g h th e di^lcritic sign s a n d \" m o r p h o p h o n e m e s \" -^just as in IP. W h a t a b o u t \" W o r d a n d P a r a d ig m \" ( W P ) ? T h e fa c t is th a t i f T W O L w o u ld a lso b e flex ib le en o u g h t o p e rm it d e s c r ip tio n s o f ty p e W P , th is w o u ld im p r o v e th e m o d e l in term s o f p e r s p ic u ity t o a c o n s id e r a b le e x te n t. I m e n tio n e d in th e p re v io u s s e ctio n th a t fo r th e sake o f th e la tte r, d ia c r itic ch a ra cte rs a re b e st a v o id e d . H ow ev er, in elim in a tin g th e \" m o r p h e m e b o u n d a r y \" (e .g . + o r -) , I h a d t o s p e cify th a t th e fron t v ow els th e su ffixes sta rt w ith d o n o t b e lo n g t o th e ste m a n d I d id th a t in u su al m a n n er-b y u sin g u p p e r ca s e letters (i.e . I a n d E ), w h ich u n d er th e p o p u la r te r m in o lo g y u su a lly g o as \" m o r p h o p h o n e m e s \" . N o w as fo r e x a m p le N y m a n (1 9 8 8 ) p o in ts o u t, th e se a re n o t in tern a l t o th e m o d e l in a n y th e o re tica l a s p e c t, b u t sim p ly e x p ress a co n v e n ie n t w a y o f e n c o d in g m o r p h o lo g ica l in fo r m a tio n in seg m en ts (e .g . I = i + PL e t c .) a n d a re n ecessa ry b e c a u s e th e tw o -le v e l ru les, o r ra th er th e fin ite -s ta te tra n sd u ce rs th e y a re c o m p ile d in, o p e r a te o n ly o n segm en ts. N o w lo o k in g a t (5 ) a b o v e , w e ca n fu rth e rm o re see th a t th is in fo r m a tio n is red u n d a n t, sin ce I is sp e cifie d as a p lu ra l su ffix in th e le x ic o n as w ell. F u rth e r m o re th ere m u st b e an \" o rd in a ry i\" p lu ra l su ffix fo r a d je c tiv e s su ch as \" p la x \" (frig h tfu l), w h ich d o n o t u n d e r g o p a la ta liz a tio n in p lu ra l fo r m ( p l a x / p l a x i ) . A p o ssib le w a y t o p reserv e th e efficie n cy th a t p r o c e s s in g s eg m en ta l rep resen ta tio n p ro v id e s, w h ile a v o id in g in co n siste n cie s a n d im p r o v in g cla r ity w o u ld b e to in c o r p o r a te a W P elem en t in th e T W O L fo rm a lism . T h is c o u ld p o s s ib ly lo o k th e fo llo w in g w ay: T h e id e a is th a t ' H O T D O G ' ca n b e d efin ed (fo r e x a m p le u n d er th e h ea d in g P A R A D I G M S ) as a p r o t o t y p ic a l en try fo r a p a ra d ig m w h ich u n d erg oes an a lte r n a tio n -it n eed n o t b e sp e cifie d w h ich , th e in fo rm a tio n fo r th is is con ta in ed in th e ru le-in p l u r a l f o r m . N o w th e sim p lest w a y to im p le m e n t th is w ou ld b e s o m e th in g like: (7) T R A N S FORM(entry (continuation)) IF (num (continuation)) = p i O n th e o th e r h a n d th e set w h ich th e p a la ta liza tio n ru le referred t o c o u ld be re d e fin e d , s o th a t it says in stea d : ( 8 ) Fron t V = i + p i , e + p i . N o w i f th e se c a n b e c o m p ile d to g e th e r , th en o n e sh o u ld b e a b le t o (th o u g h I h a v e n ' t fig u re d o u t a g en era l m ech a n ism y e t) a sso cia te th e va lu e o f T R A N S F O R M w ith th e v alu e o f r e s p e c tiv e ly i + p i a n d e + p i . A triv ia l w a y t o d o th is w o u ld b e t o ex p re ss th e first a fte r co m p ila tio n as I , th e s e c o n d as E and T R A N S F O R M w o u ld th en o n ly h ave t o b e a fu n c tio n su ch as U P P E R C A S E . In th is w a y th e resu lt w o u ld b e th e sa m e as w ith \" m o r p h o p h o n e m e s \" b u t o n a level th a t m o r e c le a r ly b e lo n g s t o th e \" m a ch in e se\" (c f. N y m a n 1 988) th a n th e present fo r m a lis m . T h e m a jo r g a in fr o m su ch a s tr a te g y w o u ld b e th a t o n e w o u ld n ot h a v e t o w o r r y a b o u t u sin g u p p e r c a s e le tte rs \" in th e rig h t p o s itio n s \" , o r t o assign d ifferen t c o n tin u a tio n cla sses fo r th e sa m e su ffix-as lo n g as o n e g ives th e right \" p a r a d ig m \" , w h ich seem s in tu itiv e ly a m u ch easier th in g to d o a n d , at th e sam e tim e , is m o r e lin g u is tica lly m o tiv a te d . I f th is, o r s o m e sim ila r m e ch a n ism , e x is te d a t lea st as an o p tio n in T W O L (w h ic h as y e t it d o e s n o t ), th en m o s t p r o b a b ly th e fle x ib ility o f th e ov e ra ll sy stem w o u ld b e in cre a se d in a w a y w h ich w o u ld a lso p e rm it faster d e v e lo p m e n t. T h is c o u ld b e d e s c r ib e d as a m a tte r o f e c o n o m y , w h ich m a y a lso b e co n sid e re d as a g o a l in itself, a lb e it m a in ly fo r p r a c tic a l sy stem s.",
                "type_str": "figure"
            },
            "FIGREF19": {
                "uris": null,
                "num": null,
                "text": "y \" e c o n o m y \" w e m ea n th a t b o t h th e c o n s tr u c tio n a n d th e ru n n in g o f a system sh o u ld n o t b e t o o e x p e n s iv e in term s o f tim e a n d w ork th en o n e c o u ld say th a t all th e c r ite r ia I h a v e d iscu ssed in th e p re v io u s su b -s e ctio n s , i.e. su fficien t gen erality, p e rsp icu ity , m o d u la r ity e tc . c a n b e re g a rd e d as relevan t. O n e m a y fu rth er p oin t o u t th a t fo r th e sak e o f e c o n o m y is is m u ch m o re p r a c tic a l to h ave a sy stem th a t an alyses a n d gen era te s ( i f b o th are req u ired , o f c o u r s e ) w ith th e sa m e p ro g ra m , th an t o u se tw o d ifferen t o n e s, a n d th is ca n o n ly b e a ch ie v e d i f th e N L P fra m e w ork is bid irection al. In th is re sp e ct, t o o , T W O L \" s cores a p o in t\" , b e c a u s e th e fin ite -s ta te tra n s d u cers are a c tu a lly c o r r e s p o n d e n c e s , n o t tra n s fo rm a tio n s , a n d it is as e a sy t o g o fr o m lex ica l t o su rfa ce fo r m , as in th e o p p o s it e d ire ctio n . O n e d ra w b a c k in all im p le m e n ta tio n s u p t o n o w , as I m e n tio n e d a b o v e , is th a t th e le x ic o n has o n ly b e en e q u ip p e d fo r r e c o g n itio n , b u t th is sh o u ld b e a m en d a b le .",
                "type_str": "figure"
            },
            "FIGREF20": {
                "uris": null,
                "num": null,
                "text": "in a lly I c o m e t o th e co n tro v e rsia l m a tte r o f w h a t it m ea n s fo r a m o d e l t o b e \" p s y c h o lo g ic a lly p la u sib le \" , a n d sin ce s o m u ch has b e e n w ritte n a n d sa id o n th e m a tte r, (e .g . L in ell 1 979) I w ill tr y t o sa y as little as p o s s ib le . E x p re ss e d w ith p r e ca u tio n , a m o d e l m a y b e reg a rd ed as p la u sib le i f it a d h e re s-o n s o m e fea sib le level-t o th e p s y c h o lo g ic a l e v id e n c e w e h a v e o f h u m a n v e rb a l b e h a v io u r. O b v io u s ly p s y c h o lo g ic a l p la u sib ility , as a g o a l fo r an N L P s y s te m , is q u a li ta tiv e ly d ifferen t fr o m th e o th e r g o a ls d iscu ssed a b o v e . F ir s tly it c o n c e r n s o n ly sy stem s w h ich e x p lic itly tr y t o m o d e l h u m an v e rb a l b e h a v io u r . A g rea t m a n y o f th em -o p e n ly o r n o t-d o n o t a sp ire t o d o s o , b e it fo r p r a c tic a l o r th e o re ti ca l reason s ( \" c o m p e te n c e m o d e ls \" e t c .). T h e n it seem s th a t su ch a g o a l w o u ld take us b a ck t o th e v a gu en ess th a t o fte n ch a ra cte rise s a p p r o a c h e s t o th e o r e tic a l lin gu istics a n d th a t w e in d ir e c tly trie d t o a v o id b y s e ttin g u p all \" g o a ls \" an d \" criteria \" u p t o n ow . T h is is n o t q u ite so. W h e n I sa id at th e b e g in n in g th a t I w as s c e p tic a l to su ch m a x im u m g o a ls as e x p la n a to r y a d e q u a c y a n d p s y c h o lo g ic a l rea lity in th e field o f co m p u ta tio n a l lin gu istics, I trie d t o stress-\" t o b e g in w ith \" . T h e p o in t o f la y in g d o w n a ltern a tiv e g o a ls w as ra th e r t o g iv e an id e a o f w h a t p r o p e r tie s form a lism s o r m o d e ls sh o u ld h ave in o r d e r t o g o b e y o n d th e lim ita tio n s in h eren t in m o st cu rren t c o m p u ta tio n a l w o rk w ith n a tu ra l la n g u a g e . W h e n th e N L P fra m ew ork s h ave b e e n d e v e lo p e d to th e d eg re e th a t it b e c o m e s m ea n in g fu l to ask q u estion s p e rta in in g m a tters su ch as p la u sib ility a n d e x p la n a to r y a d e q u a c yth en o f c o u rs e o n e s h o u ld ask th em . T W O L seem s t o h ave b e e n a b re a k th ro u g h fo r c o m p u ta tio n a l m o r p h o lo g y in this re sp e ct as w ell. O n e ca n still b e cr itic a l to w a rd s it-fo r e x a m p le c o n cern in g its stress o n eflBciency w h ile so m e w h a t n e g le ctin g p ersp icu ity , b u t o n th e w h o le -as th e d iscu ssio n o f p r o p e r tie s 3 .1 -3 .5 in d ir e c tly d e m o n s tr a te d -it p ro v id es relia b le g ro u n d t o b u ild on . P r o b a b ly it has a lso b e e n th e first c o m p u ta tio n a l m o d e l o f m o r p h o lo g y t o m ake an in terestin g h y p o th e s is a b o u t h u m a n p r o c e s s in g o f m o r p h o lo g ic a l in fo r m a tion . T h e e v id e n c e th a t K o s k e n n ie m i (1 9 8 3 ) d iscu sses is b a se d o n p e r fo r m a n c e errors. It c o n ce rn s th e fa c t th a t spea k ers-m a in ly ch ild re n a n d a p h a sics-te n d t o m ake m istak es in th a t th e y p r o d u c e w o r d -fo r m s o f m o r e p r o d u c tiv e in flection a l p a ttern s in stea d o f less p r o d u c tiv e o n e s. B u t e rro rs th a t c o r r e s p o n d t o \" a u to m a tic a lte rn a tio n s\" , i.e. p h o n o lo g ic a lly m o tiv a te d o n e s, a re e x tr e m e ly rare. S o if o n e d e scrib e s o n ly th is ty p e o f a lte rn a tio n s w ith tw o -le v e l ru les, th e fo llo w in g \" e r r o r m o d e l\" ca n b e s ta te d : \" T h e p e r fo r m a n c e e rrors in p r o d u c in g w o rd form s a re m o s t ly fa u lty c h o ic e s in th e c o n s tr u c tio n o f th e le x ica l level-th ere a re h a rd ly a n y e rro rs in th e a p p lic a tio n o f th e tw o -le v e l ru les\" (K o s k e n n ie m i 1 9 8 3 :1 3 2 ). T h e s im p lic ity a n d d ire ctn e ss o f th e tw o -le v e l ru les in c o n tra s t w ith th e in term ed ia te sta g e s a n d o rd e r in g s o f g e n e ra tiv e p h o n o lo g y rules m a k e it p o ssib le t o say th at th e T W O L fr a m e w o r k a c tu a lly p r e d i c t s th is error m o d e l. In th is sen se T W O L",
                "type_str": "figure"
            },
            "FIGREF21": {
                "uris": null,
                "num": null,
                "text": "g o a ls fo r n a tu ra l la n g u a g e p r o c e s s in g sy stem s. T o m a k e th is less sp e cu la tiv e I h a v e c o n c e n tr a te d o n c o m p u ta tio n a l m o r p h o lo g y an d m o st o f all o n o n e sin g le rep resen ta tiv e-th e tw o -le v e l m o d e l. T h e e x a m p le s o f B u lg a ria n m o r p h o lo g y h a v e b e e n p r o v id e d in th e b e lie f th a t \" m e ta th e o r e tic a l\" stu d ie s in lin gu istics can n o t b e c a rrie d o n s u cce ssfu lly w ith o u t \" a n ch o rin g s\" in p a rticu la r la n g u a g es. T h e c h o ic e o f B u lg a ria n has b e e n d ic t a te d b y th e fa c t th a t o n ly th ere cein I claim s o m e o rig in a lity . T h e g o a ls w h ich I h a v e b r o u g h t u p a re o f co u rse n o t a b so lu te , b u t I h a v e held th a t th e y p r o v id e a g o o d s ta r tin g p o in t fo r s e ttin g \" sta n d a rd s\" in co m p u ta tio n a l lin g u istics a n d m o r e s p e c ific a lly -in c o m p u ta tio n a l m o rp h o lo g y . F or e a ch o f th ese g o a ls I h a v e d iscu ssed p r o p e r tie s o f th e fra m e w ork s th a t th e sy stem s are b a sed o n , w h ich a re n ece ssa ry o r s im p ly d e sira b le fo r th e sake o f th ese g o a ls. T h e fo llo w in g ta b le sh ow s th e se o n c e m o re , G O A L S o f N L P S Y S T E M S P R O P E R T I E S o f N L P F R A M E W O R k c o m p le te n e ss R e strictiv e n e ss P e r s p ic u ity D e cla ra tiv e n e ss M o d u la r it y H a rd -a n d softw a re in d e p e n d e n ce L in g u is tic fle x ib ility B id ir e c tio n a lity P s y c h o lo g ic a l p la u s ib ility ? It is th e g o a ls th a t are th e m o r e s ta b le p a r t. D ifferen t sy stem s ca n la y stress o n s o m e as o p p o s e d t o o th e r s b u t still s y s te m -g o a ls are less co n tro v e rsia l a n d less c o n t r a d ic t o r y t o o n e a n o th e r . T h e p r o p e r tie s , h ow e v er, ca n in te ra ct in c o m p le x w a y s a n d s o m e tim e s w o r k a g a in s t s o m e o th e r g o a l th a n th e o n e w h ich has ca lled fo r th e m in th e first p la c e . S o th ere is n o e x a c t re la tion b etw een g o a ls an d p r o p e r tie s , b u t ra th er a d y n a m ic p r o c e s s g iv in g p r io r ity t o s o m e in stea d o f o th ers in p a r tic u la r ca ses. D e s p ite th a t, I b e lie v e th a t th e se p r o p e r tie s ca n b e reg a rd e d as g u id elin es fo r w o r k in th e field o f c o m p u ta tio n a l m o r p h o lo g y a n d p r o b a b ly in c o m p u ta tio n a l lin gu istics as a w h ole, w h ich is ev en m o r e im p o r ta n t in th e a b sen se o f cle a r c r it e ria fo r w h a t sh o u ld b e co n s id e r e d as p s y c h o lo g ic a l p la u s ib ility a n d e x p la n a to r y a d e q u a c y in th e field o f th e o r e tic a l lin g u istics a n d th e n e ig b o u r in g d iscip lin e s.T h e q u e stion m a rk in th e ta b le a b o v e s h o u ld sig n ify th a t. W h a te v e r its an sw er, I",
                "type_str": "figure"
            },
            "FIGREF22": {
                "uris": null,
                "num": null,
                "text": "lle d (in 3 .4 .) \" lin g u istic fle x ib ility \" , i.e. th e p o s s ib ility o f e x pressin g la n g u a g e d e s c r ip tio n s a c c o r d in g t o differen t ty p e s o f th e o r e tic a l m o d e ls . H ow ever, I see th is as a m ea n s fo r a ch ie v in g th e a im , n o t as th e aim itself. T h e w h o le p o in t o f w o rk in g in d e p e n d e n tly o f th e o r y is o n ly t o re tu rn t o th e ory, b u t w ith less secta ria n ism an d g r e a te r in sig h t. O n ly th is w a y w ill c o m p u t a tio n a l lin g u istics c o n tr ib u te t o th e u n d e rs ta n d in g o f w h a t a p s y c h o lo g ic a lly p la u sib le m o d e l sh o u ld b e , w h ich is a p re re q u isite fo r a p p r o 2u:bing a m o d e l /t h e o r y o f n a tu ra l la n g u a g e th a t is \" p s y c h o lo g ic a lly rea l\" . In d o in g s o , c o m p u ta tio n a l lin g u istics is b o u n d t o c o m e closer a n d p r o b a b ly m erg e w ith th e b r o a d e r p a ra d ig m o f c o g n itiv e scien ce.",
                "type_str": "figure"
            }
        }
    }
}