File size: 47,432 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
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
{
    "paper_id": "2019",
    "header": {
        "generated_with": "S2ORC 1.0.0",
        "date_generated": "2023-01-19T07:29:19.208839Z"
    },
    "title": "A Metric for Lexical Complexity in Malayalam",
    "authors": [
        {
            "first": "Richard",
            "middle": [],
            "last": "Shallam",
            "suffix": "",
            "affiliation": {},
            "email": ""
        },
        {
            "first": "Ashwini",
            "middle": [],
            "last": "Vaidya",
            "suffix": "",
            "affiliation": {},
            "email": "avaidya@hss.iitd.ac.in"
        }
    ],
    "year": "",
    "venue": null,
    "identifiers": {},
    "abstract": "This paper proposes a metric to quantify lexical complexity in Malayalam. The metric utilizes word frequency, orthography and morphology as the three factors affecting visual word recognition in Malayalam. Malayalam differs from other Indian languages due to its agglutinative morphology and orthography, which are incorporated into our model. The predictions made by our model are then evaluated against reaction times in a lexical decision task. We find that reaction times are predicted by frequency, morphological complexity and script complexity. We also explore the interactions between morphological complexity with frequency and script in our results. To the best of our knowledge, this is the first study on lexical complexity in Malayalam.",
    "pdf_parse": {
        "paper_id": "2019",
        "_pdf_hash": "",
        "abstract": [
            {
                "text": "This paper proposes a metric to quantify lexical complexity in Malayalam. The metric utilizes word frequency, orthography and morphology as the three factors affecting visual word recognition in Malayalam. Malayalam differs from other Indian languages due to its agglutinative morphology and orthography, which are incorporated into our model. The predictions made by our model are then evaluated against reaction times in a lexical decision task. We find that reaction times are predicted by frequency, morphological complexity and script complexity. We also explore the interactions between morphological complexity with frequency and script in our results. To the best of our knowledge, this is the first study on lexical complexity in Malayalam.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Abstract",
                "sec_num": null
            }
        ],
        "body_text": [
            {
                "text": "The task of visual word recognition is related to language processing at the level of a word/lexical item. A word can be analyzed at several linguistic levels, and the word recognition task helps us understand the role of these levels in relation to processing, memory and attention. In psycholinguistics, previous work on this topic focuses on understanding the individual variables that affect the lexical processing of words. If we can quantify the influence of variables ranging from orthographic features to semantic factors on the cognitive processing of words, it would help us in understanding the critical factors underlying visual word recognition (and pattern recognition, more generally). The resulting model of word recognition can be evaluated against human judgements.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Introduction",
                "sec_num": "1"
            },
            {
                "text": "Models of word recognition are especially relevant for eye-tracking studies, where they have been extensively explored (Rayner and Duffy, 1986) . Word recognition models have also been used to understand reading disabilities such as phonological and surface dyslexia (Balota et al., 2006) . For these studies, it is crucial to tease apart the effect of various factors that affect the task of reading. Previous research has shown that the eye gaze duration is affected by frequency, orthography, morphology and phonology, among others. Apart from these studies, an understanding of lexical complexity is also an interesting topic for study on its own.",
                "cite_spans": [
                    {
                        "start": 119,
                        "end": 143,
                        "text": "(Rayner and Duffy, 1986)",
                        "ref_id": "BIBREF6"
                    },
                    {
                        "start": 267,
                        "end": 288,
                        "text": "(Balota et al., 2006)",
                        "ref_id": "BIBREF0"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Introduction",
                "sec_num": "1"
            },
            {
                "text": "In this paper, we explore the case of Malayalam and in particular examine three factors that could predict word complexity in the language: frequency, orthography and morphology. The role of variables that determine word recognition in Malayalam has not been explored, as it has been for Hindi (Husain et al., 2015; Verma et al., 2018) . Quantifying these factors in a model of lexical complexity can help us in developing norms that are useful in areas such as reading studies and word generation for lexical decision tasks. Further, this would contribute towards cross-linguistic comparison of these factors from a different language family. To the best of our knowledge, this is the first work that examines lexical complexity in Malayalam.",
                "cite_spans": [
                    {
                        "start": 294,
                        "end": 315,
                        "text": "(Husain et al., 2015;",
                        "ref_id": "BIBREF4"
                    },
                    {
                        "start": 316,
                        "end": 335,
                        "text": "Verma et al., 2018)",
                        "ref_id": "BIBREF9"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Introduction",
                "sec_num": "1"
            },
            {
                "text": "The task of visual word recognition involves the cognitive processing of visual information and comparing it with a particular internal mental representation of a word. This representation itself may be at the graphemic, phonemic, morphemic and lexical semantic level, all of which have been shown to affect word recognition (Balota et al., 2006) . In the sections that follow, we describe the three factors that are included in our study.",
                "cite_spans": [
                    {
                        "start": 325,
                        "end": 346,
                        "text": "(Balota et al., 2006)",
                        "ref_id": "BIBREF0"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Lexical Complexity",
                "sec_num": "2"
            },
            {
                "text": "The effect of word frequency is robust and has been well studied across word recognition tasks (Balota et al., 2006) . High frequency words tend to be recognized faster than low frequency words. In eye tracking studies high frequency words have lower gaze duration and fixation measures. We would expect that frequency would have a similar effect on the Malayalam data, where high frequency would contribute towards a lower lexical complexity.",
                "cite_spans": [
                    {
                        "start": 95,
                        "end": 116,
                        "text": "(Balota et al., 2006)",
                        "ref_id": "BIBREF0"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Word Frequency",
                "sec_num": "2.1"
            },
            {
                "text": "A word may be composed of a single morpheme e.g. boy or more than one e.g. funnily: funny+ ly. The role of morphology in word recognition is at a sub-lexical level. Morphology as a measure is particularly relevant for an agglutinative language such as Malayalam, which also exhibits productive word compounding e.g. Just the word \u0d2e\u0d30\u0d02 (mara) \"tree\" has a number of morphological forms such as \u0d2e\u0d30 \u0d3f\u0d7d (marattil) -in the tree \u0d2e\u0d30 \u0d3f\u0d46 (marattinr\u0331 e) -of the tree \u0d2e\u0d30 \u0d7e \u0d3f\u0d1f\u0d2f\u0d3f \u0d46\u0d1f (mara\u1e45\u1e45a\u1e37kki\u1e6dayil\u016b\u1e6de) -through the trees \u0d2e\u0d30\u0d46 \u0d3e \u0d15\u0d7e (marakke\u0101mpuka\u1e37) -tree branches Early studies that looked at the effect of morphology on lexical access have suggested that polymorphemic words (i.e. words consisting of more than one morpheme) are decomposed into their component parts during online processing. This process would find the root first (e.g. funny and on finding it, proceed to search stored affix-stem combinations till funnily is retrieved (Taft and Forster, 1975) . In a morphologically-rich language such as Malayalam, we would expect that this would be an important factor in lexical processing.",
                "cite_spans": [
                    {
                        "start": 926,
                        "end": 950,
                        "text": "(Taft and Forster, 1975)",
                        "ref_id": "BIBREF7"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Morphology",
                "sec_num": "2.2"
            },
            {
                "text": "The visual processing of words involves processing at the orthographic level as well. This implies that the writing system of various languages will influence recognition. A writing system-whether alpha-syllabic, logographic or alphabetic has been shown to influence reading times (Katz and Frost, 1992) . Sub-lexical properties such as letter features and their interactions with the words themselves can also influence word complexity, which needs to be accounted for in the model.",
                "cite_spans": [
                    {
                        "start": 281,
                        "end": 303,
                        "text": "(Katz and Frost, 1992)",
                        "ref_id": "BIBREF5"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Orthography",
                "sec_num": "2.3"
            },
            {
                "text": "In order to compute the lexical complexity metric, token frequency, morphology and orthography were included as our variables. Below, the methods for computing the values for each of these variables are discussed.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Method",
                "sec_num": "3"
            },
            {
                "text": "In order to compute our metric for Malayalam, we first obtained a corpus from the Leipzig Corpora Collection containing 300,000 sentences from Malayalam Wikipedia articles and 100,000 sentences from Malayalam news crawl (Goldhahn et al., 2012) . The corpus was then preprocessed by removing punctuation and special characters, and then tokenized using whitespace. The text was also normalized to remove inconsistencies in spelling using the Indic NLP Library 1 and this resulted in 4,711,219 tokens and 762,858 unique types.",
                "cite_spans": [
                    {
                        "start": 220,
                        "end": 243,
                        "text": "(Goldhahn et al., 2012)",
                        "ref_id": "BIBREF3"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Corpus",
                "sec_num": "3.1"
            },
            {
                "text": "The corpus was used to collect counts for each word and then scaled them between 0 and 1, which was then inverted such that the most frequent tokens have a value closer to 0 and the less frequent tokens will have a value approaching 1. This score indicated the relative frequency of each word in this corpus, and the idea that highly frequent words are much easier to process than those that have lower frequency.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Word Frequency Metric",
                "sec_num": "3.2"
            },
            {
                "text": "Our morphology metric required us to obtain information about the root and the morpho-logical affixes for a given word. Given the rich morphology and compounding processes in the language, we had to make use of a two-step process to compute our scores.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Morphology Metric",
                "sec_num": "3.3"
            },
            {
                "text": "First, SandhiSplitter (Devadath et al., 2014) was used to split tokens that are compound words into their constituent component words. For example, consider the compound word \u0d15\u0d3e\u0d30\u0d23\u0d2e\u0d3e\u0d2f\u0d3f\u0d30\u0d3f \u0d23\u0d02 (kAraNamAyirikkaNaM) \u0d15\u0d3e\u0d30\u0d23\u0d2e\u0d3e\u0d2f\u0d3f\u0d30\u0d3f \u0d23\u0d02 \u21d2 \u0d15\u0d3e\u0d30\u0d23\u0d02 + \u0d06\u0d2f\u0d3f\u0d30\u0d3f \u0d23\u0d02 k\u0101ra\u1e47am\u0101yirikka\u1e47am\u0307\u21d2 k\u0101ra\u1e47am\u0307+ \u0101yirikka\u1e47am\" must be the reason\" \u21d2 \"reason\" + \"must be\"",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Morphology Metric",
                "sec_num": "3.3"
            },
            {
                "text": "As a second step, these results were passed through IndicStemmer 2 , a rule-based stemmer for Malayalam, which further decomposed the words into stems and affixes. As an example, the word \u0d47\u0d32\u0d16\u0d28 \u0d46\u0d1f (l\u0113khana\u1e45\u1e45a\u1e37u\u1e6de) meaning \"Of articles\". is decomposed into the stem \u0d47\u0d32\u0d16\u0d28\u0d02 (l\u0113khanam) meaning article with the suffix -\u0d7e ( \u1e45\u1e45al) indicating plural and --\u0d41\u0d46\u0d1f (u\u1e6de) indicating the Genitive case. In our metric we only considered suffixes as in Malayalam usually contains always suffixes being added to the end of the stem.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Morphology Metric",
                "sec_num": "3.3"
            },
            {
                "text": "After this two-step process, we are able to obtain the stems and suffixes for a given word.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Morphology Metric",
                "sec_num": "3.3"
            },
            {
                "text": "By simply summing the number of stems and suffixes, the total number of morphemes contained in each word is computed. For example, the word \u0d38 \u0d3f \u0d02 (sampatsamr d'dhiyum) meaning \"prosperity\" is a compound word split into constituent words \u0d38 \u0d4d (sampatt) meaning \"richness\" and \u0d38 \u0d3f \u0d02 (samr d'dhiyum) meaning \"and plentiful\". \u0d38 \u0d3f \u0d02 (samr d'dhiyum) is further stemmed to stem word \u0d38 \u0d3f (samr d'dhi) meaning \"plentiful\" and suffix -\u0d41\u0d02 (um) meaning \"-and\". \u0d38 \u0d4d (sampatt) is a root word. Thus, the number of morphemes in this case is three, counting the two stems and one suffix.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Morpheme Count",
                "sec_num": null
            },
            {
                "text": "Based on this pre-processing, we then calculate the total number of morphemes for each whole word and then scale this number between 0 and 1 to give a morpheme score. We note that there could be several different ways to compute the morpheme score, as affixes themselves are not all alike. In this preliminary study, it was not immediately apparent how the differing costs for various affixes could be calculated. Additionally, fine-grained information regarding the morphological properties of the affixes (e.g. whether they were inflectional or derivational) was not easily obtained with existing tools and resources. In future work, we plan to explore this possibility by enhancing the morphological analyzer's output.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Morpheme Count",
                "sec_num": null
            },
            {
                "text": "Malayalam is an alphasyllabic writing system that has its source in the Vatteluttu alphabet from the 9 th century. Its modern alphabets have been borrowed from the Grantha alphabet. It consists of 15 vowels and 36 consonant letters.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Orthography Metric",
                "sec_num": "3.4"
            },
            {
                "text": "We devised a script score based on complexity of the script in the following three ways:-",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Orthography Metric",
                "sec_num": "3.4"
            },
            {
                "text": "In the alpha-syllabic script of Malayalam, vowels may either appear as letters at the beginning of a word or as diacritics. Consonants themselves are understood to have an inherent schwa, which is not separately represented. The diacritics will appear either left or right of the consonant it modifies. If it appears to the left, there will be a discrepancy in the phonemic and the orthographic order, as the vowel will always be pronounced after the consonant, but read before the consonant actually appear in the text. For example:",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Mismatch in Spoken and Visual Order",
                "sec_num": null
            },
            {
                "text": "\u0d15 +\u0d46\u25cc = \u0d46\u0d15 ka + .e = ke",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Mismatch in Spoken and Visual Order",
                "sec_num": null
            },
            {
                "text": "Here the vowel violates the order in which it is spoken. Similarly: \u0d15 +\u0d47\u25cc = \u0d47\u0d15 (ka + \u0113 = k\u0113), as seen in \u0d47\u0d15\u0d7e \u0d15 (k\u0113\u1e37kkuka) meaning \"hear\". Such inconsistencies in spoken and visual order have been shown to incur a cost in Hindi word recognition (which is also an alpha-syllabic script) (Vaid and Gupta, 2002) .",
                "cite_spans": [
                    {
                        "start": 285,
                        "end": 307,
                        "text": "(Vaid and Gupta, 2002)",
                        "ref_id": "BIBREF8"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Mismatch in Spoken and Visual Order",
                "sec_num": null
            },
            {
                "text": "In order to capture the lexical processing cost for such a discrepancy, we give a penalty of 1 every time it occurs in the word.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Mismatch in Spoken and Visual Order",
                "sec_num": null
            },
            {
                "text": "In Malayalam, the diacritic may also appear above or below a consonant. In such a case, we we give a penalty of 0.5 to the word. For example the symbol \u25cc\u0d4d also known as virama is used to replace the inherent schwa sound of consonants with \u016d. As in \u0d15 + \u25cc\u0d4d = \u0d15\u0d4d (ka + virama = ku)",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Diacritic Appearing Above or Below",
                "sec_num": null
            },
            {
                "text": "A penalty of one is assigned for every two letters that form a composite glyph. For example: \u0d2e \u0d3f (mantri) = \u0d2e\u0d28\u0d4d + \u0d3f (man + tri) where the new composite glyph is (ntra). With the above complexity rules in place, the total penalty cost for each whole word is calculated. Then the total penalty for each word is scaled linearly to between 0 and 1 to give us an orthographic score.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Ligatures and Consonant Clusters",
                "sec_num": null
            },
            {
                "text": "In order to evaluate our lexical complexity metric, we used a lexical decision task paradigm to collect reaction times for a sample of Malayalam words. More complex words would result in longer reaction times, and vice versa. This would help us evaluate whether our lexical complexity model could predict reaction times for the given set of words. We used a well-understood experimental paradigm in the form of a lexical decision task. In such a setup, a participant will see a word stimuli on a screen which they have to classify as either a word or a non-word using a button press. The response time (RT) is calculated from the point the word appears on the screen to the point where the participant presses the response button.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Evaluation of the Complexity Metric",
                "sec_num": "3.5"
            },
            {
                "text": "Our task consisted of a balanced set of 50 Malayalam words and 50 pseudowords. Pseudowords follow the phonotactics of the language, but have no lexical meaning (i.e. are not legitimate words). In order to select words for the task, two sets of 25 words were randomly sampled from the unique tokens obtained from the Leipzig Corpus. The first set was randomly sampled from words with a frequency score between the range of 0.1 to Figure 1 : Stimuli word shown for 2500ms. The first word is a proper Malayalam word (\"vivara\u1e45\u1e45a\u1e37\" meaning \"information\") hence the correct response is to press the 'a' key. The second word is non-word (vamittam) and therefore, the correct response is to press 'l' key.",
                "cite_spans": [],
                "ref_spans": [
                    {
                        "start": 429,
                        "end": 437,
                        "text": "Figure 1",
                        "ref_id": null
                    }
                ],
                "eq_spans": [],
                "section": "Materials",
                "sec_num": null
            },
            {
                "text": "0.4 to obtain high frequency words as calculated by the metric. The second set was chosen similarly but with frequency score between the range of 0.7 to 0.9 to yield low frequency words. If the sampled word turned out to be an English word written in Malayalam or happens to be a proper noun, it was replaced with another until both sets had 25 words each.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Materials",
                "sec_num": null
            },
            {
                "text": "The pseudowords were constructed in keeping with the phonotactics of Malayalam. Both the pseudowords and the valid words were constrained in length between 6 and 14 characters. Note that we do not take into consideration the reaction times for the pseudowords; they are simply distractors for the participants.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Materials",
                "sec_num": null
            },
            {
                "text": "Participants included 38 students from S.N. College, Kerala, who volunteered for the study. Participants included 20 females and 18 males between the ages of 18 and 23 (mean age of 19.7). All participants were native speakers of Malayalam and had formal education in Malayalam upto grade 10. ",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Participants",
                "sec_num": null
            },
            {
                "text": "Participants were tested individually on a computer running the lexical decision task on the JsPsych stimulus presentation software (De Leeuw, 2015) . Each participant was asked to press either the 'a' key or the 'l' key for word and non-word respectively. The order of words and pseudowords was randomized for each participant. Participants were instructed to read the word presented and respond with the appropriate button press. Each trial consisted of a word that was presented for 2500ms. A fixation cross was placed in the center for 1600ms between each trial. The first 10 trials were practice trials from a word set different from the study. This enabled participants to get familiarized with the task.",
                "cite_spans": [
                    {
                        "start": 136,
                        "end": 148,
                        "text": "Leeuw, 2015)",
                        "ref_id": "BIBREF1"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Procedure",
                "sec_num": null
            },
            {
                "text": "The trials belonging to those who scored below 70% in word-non-word accuracy were excluded, which brought the number of participants to 35.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Results",
                "sec_num": "4"
            },
            {
                "text": "We fit a linear model using the lm function in R. Log reaction times were used with frequency, script and morph as the covariates. Figure 2 shows that the three variables are not highly correlated in our test set. Table 1 shows the results of the regression analysis. The main inference we can draw from the result is that the variables Script, Morphology and Frequency have a significant effect (all p-values < 0.05) on (reaction times) RTs, such that a high cost of script, morph and frequency leads to higher RTs.",
                "cite_spans": [],
                "ref_spans": [
                    {
                        "start": 131,
                        "end": 139,
                        "text": "Figure 2",
                        "ref_id": "FIGREF0"
                    },
                    {
                        "start": 214,
                        "end": 221,
                        "text": "Table 1",
                        "ref_id": null
                    }
                ],
                "eq_spans": [],
                "section": "Results",
                "sec_num": "4"
            },
            {
                "text": "In addition, the results also indicate a Table 1 : Results for all three variables and their interactions. Script and Morphological Complexity as well as Frequency and Morphological Complexity show a significant interaction marginal interaction between Script and Morphology (p=0.06), such that an increase in the script complexity leads to larger increases in RTs for morphologically simpler words (Cost <0.9) compared to morphologically complex words (Cost >0.9) (see Figure 3 ). There is also a marginal interaction between Morphology and Frequency (p=0.08) such that an increase in the frequency cost leads to higher reaction times in morphologically complex words as compared to morphologically simpler words (see Figure 4 ). ",
                "cite_spans": [],
                "ref_spans": [
                    {
                        "start": 41,
                        "end": 48,
                        "text": "Table 1",
                        "ref_id": null
                    },
                    {
                        "start": 470,
                        "end": 478,
                        "text": "Figure 3",
                        "ref_id": "FIGREF1"
                    },
                    {
                        "start": 719,
                        "end": 727,
                        "text": "Figure 4",
                        "ref_id": "FIGREF2"
                    }
                ],
                "eq_spans": [],
                "section": "Results",
                "sec_num": "4"
            },
            {
                "text": "Our results replicate the robust effects of frequency on lexical processing in Malayalam.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Discussion",
                "sec_num": "5"
            },
            {
                "text": "As frequency is a known predictor of reaction times, we expected to find a significant effect for frequency, but we particularly wanted to understand the effect of morphology and orthography on word recognition in Malayalam. Orthographic complexity as captured by diacritic placement and ligatures also has a significant effect on lexical processing. Similarly, we also find an effect for morphological complexity in terms of the number of morphemes in a word.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Discussion",
                "sec_num": "5"
            },
            {
                "text": "The interactions in our model point to an interesting relationship between high frequency words and morphological complexity. It appears that the effect of frequency cost becomes more pronounced in more complex words. In other words, low frequency words lead to higher reaction times particularly when they are morphologically complex. Perhaps this is because the cost of lexical decomposition is higher in these words. On the other hand, the effect size of script is weaker and becomes visible only when the word is morphologically simple. When the word is morphologically complex, this effect is not very apparent.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Discussion",
                "sec_num": "5"
            },
            {
                "text": "This work points to many interesting future avenues for exploring lexical complexity in an agglutinative language like Malayalam. Particularly, the effect of morphological complexity on factors like frequency need to be explored more thoroughly. In the future, we plan to carry out experiments with a larger set of items for the lexical decision task, as this was a preliminary study. We also plan to experiment with other measures of morphological complexity that take into account information about the type as well as the number of morphemes.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Discussion",
                "sec_num": "5"
            },
            {
                "text": "https://anoopkunchukuttan.github.io/indic_ nlp_library/",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            },
            {
                "text": "https://github.com/libindic/indicstemmer",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            }
        ],
        "back_matter": [],
        "bib_entries": {
            "BIBREF0": {
                "ref_id": "b0",
                "title": "Visual word recognition: The journey from features to meaning (a travel update)",
                "authors": [
                    {
                        "first": "A",
                        "middle": [],
                        "last": "David",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Balota",
                        "suffix": ""
                    },
                    {
                        "first": "J",
                        "middle": [],
                        "last": "Melvin",
                        "suffix": ""
                    },
                    {
                        "first": "Michael J",
                        "middle": [],
                        "last": "Yap",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Cortese",
                        "suffix": ""
                    }
                ],
                "year": 2006,
                "venue": "Handbook of Psycholinguistics",
                "volume": "",
                "issue": "",
                "pages": "285--375",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "David A Balota, Melvin J Yap, and Michael J Cortese. 2006. Visual word recognition: The journey from features to meaning (a travel up- date). In Handbook of Psycholinguistics, pages 285-375. Elsevier.",
                "links": null
            },
            "BIBREF1": {
                "ref_id": "b1",
                "title": "jspsych: A javascript library for creating behavioral experiments in a web browser. Behavior research methods",
                "authors": [
                    {
                        "first": "Joshua R De",
                        "middle": [],
                        "last": "Leeuw",
                        "suffix": ""
                    }
                ],
                "year": 2015,
                "venue": "",
                "volume": "47",
                "issue": "",
                "pages": "1--12",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Joshua R De Leeuw. 2015. jspsych: A javascript library for creating behavioral experiments in a web browser. Behavior research methods, 47(1):1-12.",
                "links": null
            },
            "BIBREF2": {
                "ref_id": "b2",
                "title": "A sandhi splitter for malayalam",
                "authors": [
                    {
                        "first": "",
                        "middle": [],
                        "last": "Vv Devadath",
                        "suffix": ""
                    },
                    {
                        "first": "J",
                        "middle": [],
                        "last": "Litton",
                        "suffix": ""
                    },
                    {
                        "first": "Dipti",
                        "middle": [
                            "Misra"
                        ],
                        "last": "Kurisinkel",
                        "suffix": ""
                    },
                    {
                        "first": "Vasudeva",
                        "middle": [],
                        "last": "Sharma",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Varma",
                        "suffix": ""
                    }
                ],
                "year": 2014,
                "venue": "Proceedings of the 11th International Conference on Natural Language Processing",
                "volume": "",
                "issue": "",
                "pages": "156--161",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "VV Devadath, Litton J Kurisinkel, Dipti Misra Sharma, and Vasudeva Varma. 2014. A sandhi splitter for malayalam. In Proceedings of the 11th International Conference on Natural Lan- guage Processing, pages 156-161.",
                "links": null
            },
            "BIBREF3": {
                "ref_id": "b3",
                "title": "Building large monolingual dictionaries at the leipzig corpora collection: From 100 to 200 languages",
                "authors": [
                    {
                        "first": "Dirk",
                        "middle": [],
                        "last": "Goldhahn",
                        "suffix": ""
                    },
                    {
                        "first": "Thomas",
                        "middle": [],
                        "last": "Eckart",
                        "suffix": ""
                    },
                    {
                        "first": "Uwe",
                        "middle": [],
                        "last": "Quasthoff",
                        "suffix": ""
                    }
                ],
                "year": 2012,
                "venue": "LREC",
                "volume": "29",
                "issue": "",
                "pages": "31--43",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Dirk Goldhahn, Thomas Eckart, and Uwe Quasthoff. 2012. Building large monolingual dictionaries at the leipzig corpora collection: From 100 to 200 languages. In LREC, vol- ume 29, pages 31-43.",
                "links": null
            },
            "BIBREF4": {
                "ref_id": "b4",
                "title": "Integration and prediction difficulty in hindi sentence comprehension: Evidence from an eye-tracking corpus",
                "authors": [
                    {
                        "first": "Samar",
                        "middle": [],
                        "last": "Husain",
                        "suffix": ""
                    },
                    {
                        "first": "Shravan",
                        "middle": [],
                        "last": "Vasishth",
                        "suffix": ""
                    },
                    {
                        "first": "Narayanan",
                        "middle": [],
                        "last": "Srinivasan",
                        "suffix": ""
                    }
                ],
                "year": 2015,
                "venue": "Journal of Eye Movement Research",
                "volume": "8",
                "issue": "2",
                "pages": "1--12",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Samar Husain, Shravan Vasishth, and Narayanan Srinivasan. 2015. Integration and prediction dif- ficulty in hindi sentence comprehension: Evi- dence from an eye-tracking corpus. Journal of Eye Movement Research, 8(2):1-12.",
                "links": null
            },
            "BIBREF5": {
                "ref_id": "b5",
                "title": "The reading process is different for different orthographies: The orthographic depth hypothesis",
                "authors": [
                    {
                        "first": "Leonard",
                        "middle": [],
                        "last": "Katz",
                        "suffix": ""
                    },
                    {
                        "first": "Ram",
                        "middle": [],
                        "last": "Frost",
                        "suffix": ""
                    }
                ],
                "year": 1992,
                "venue": "Advances in psychology",
                "volume": "94",
                "issue": "",
                "pages": "67--84",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Leonard Katz and Ram Frost. 1992. The read- ing process is different for different orthogra- phies: The orthographic depth hypothesis. In Advances in psychology, volume 94, pages 67- 84. Elsevier.",
                "links": null
            },
            "BIBREF6": {
                "ref_id": "b6",
                "title": "Lexical complexity and fixation times in reading: Effects of word frequency, verb complexity, and lexical ambiguity",
                "authors": [
                    {
                        "first": "Keith",
                        "middle": [],
                        "last": "Rayner",
                        "suffix": ""
                    },
                    {
                        "first": "Susan",
                        "middle": [
                            "A"
                        ],
                        "last": "Duffy",
                        "suffix": ""
                    }
                ],
                "year": 1986,
                "venue": "Memory & cognition",
                "volume": "14",
                "issue": "3",
                "pages": "191--201",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Keith Rayner and Susan A Duffy. 1986. Lexical complexity and fixation times in reading: Effects of word frequency, verb complexity, and lexical ambiguity. Memory & cognition, 14(3):191-201.",
                "links": null
            },
            "BIBREF7": {
                "ref_id": "b7",
                "title": "Lexical storage and retrieval of prefixed words",
                "authors": [
                    {
                        "first": "Marcus",
                        "middle": [],
                        "last": "Taft",
                        "suffix": ""
                    },
                    {
                        "first": "Kenneth I",
                        "middle": [],
                        "last": "Forster",
                        "suffix": ""
                    }
                ],
                "year": 1975,
                "venue": "Journal of verbal learning and verbal behavior",
                "volume": "14",
                "issue": "6",
                "pages": "638--647",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Marcus Taft and Kenneth I Forster. 1975. Lexical storage and retrieval of prefixed words. Journal of verbal learning and verbal behavior, 14(6):638- 647.",
                "links": null
            },
            "BIBREF8": {
                "ref_id": "b8",
                "title": "Exploring word recognition in a semi-alphabetic script: The case of devanagari",
                "authors": [
                    {
                        "first": "Jyotsna",
                        "middle": [],
                        "last": "Vaid",
                        "suffix": ""
                    },
                    {
                        "first": "Ashum",
                        "middle": [],
                        "last": "Gupta",
                        "suffix": ""
                    }
                ],
                "year": 2002,
                "venue": "Brain and Language",
                "volume": "81",
                "issue": "1-3",
                "pages": "679--690",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Jyotsna Vaid and Ashum Gupta. 2002. Explor- ing word recognition in a semi-alphabetic script: The case of devanagari. Brain and Language, 81(1-3):679-690.",
                "links": null
            },
            "BIBREF9": {
                "ref_id": "b9",
                "title": "Shabd: A psycholinguistics database for hindi words",
                "authors": [
                    {
                        "first": "Ark",
                        "middle": [],
                        "last": "Verma",
                        "suffix": ""
                    },
                    {
                        "first": "V",
                        "middle": [],
                        "last": "Sikarwar",
                        "suffix": ""
                    },
                    {
                        "first": "H",
                        "middle": [],
                        "last": "Yadav",
                        "suffix": ""
                    },
                    {
                        "first": "J",
                        "middle": [],
                        "last": "Ranjith",
                        "suffix": ""
                    },
                    {
                        "first": "Pawan",
                        "middle": [],
                        "last": "Kumar",
                        "suffix": ""
                    }
                ],
                "year": 2018,
                "venue": "Proceedings of ACCS",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Ark Verma, V. Sikarwar, H. Yadav, J. Ranjith, and Pawan Kumar. 2018. Shabd: A psycholinguis- tics database for hindi words. In Proceedings of ACCS 2018.",
                "links": null
            }
        },
        "ref_entries": {
            "FIGREF0": {
                "text": "Heat plot showing correlation between the three variables in our test data",
                "uris": null,
                "num": null,
                "type_str": "figure"
            },
            "FIGREF1": {
                "text": "Interaction between Morphological Complexity and Script Complexity",
                "uris": null,
                "num": null,
                "type_str": "figure"
            },
            "FIGREF2": {
                "text": "Interaction between Morphological Complexity and Frequency Cost. Note that a low Frequency Cost corresponds to a high Frequency Count for a word",
                "uris": null,
                "num": null,
                "type_str": "figure"
            }
        }
    }
}