File size: 68,512 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
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
{
    "paper_id": "2005",
    "header": {
        "generated_with": "S2ORC 1.0.0",
        "date_generated": "2023-01-19T07:17:11.741564Z"
    },
    "title": "Low Cost Portability for Statistical Machine Translation based on N-gram Frequency and TF-IDF",
    "authors": [
        {
            "first": "Matthias",
            "middle": [],
            "last": "Eck",
            "suffix": "",
            "affiliation": {
                "laboratory": "",
                "institution": "Carnegie Mellon University Pittsburgh",
                "location": {
                    "postCode": "15213",
                    "region": "PA",
                    "country": "USA"
                }
            },
            "email": ""
        },
        {
            "first": "Stephan",
            "middle": [],
            "last": "Vogel",
            "suffix": "",
            "affiliation": {
                "laboratory": "",
                "institution": "Carnegie Mellon University Pittsburgh",
                "location": {
                    "postCode": "15213",
                    "region": "PA",
                    "country": "USA"
                }
            },
            "email": "vogel@cs.cmu.edu"
        },
        {
            "first": "Alex",
            "middle": [],
            "last": "Waibel",
            "suffix": "",
            "affiliation": {
                "laboratory": "",
                "institution": "Carnegie Mellon University Pittsburgh",
                "location": {
                    "postCode": "15213",
                    "region": "PA",
                    "country": "USA"
                }
            },
            "email": "waibel@cs.cmu.edu"
        }
    ],
    "year": "",
    "venue": null,
    "identifiers": {},
    "abstract": "Statistical machine translation relies heavily on the available training data. In some cases it is necessary to limit the amount of training data that can be created for or actually used by the systems. We introduce weighting schemes which allow us to sort sentences based on the frequency of unseen n-grams. A second approach uses TF-IDF to rank the sentences. After sorting we can select smaller training corpora and we are able to show that systems trained on much less training data achieve a very competitive performance compared to baseline systems using all available training data.",
    "pdf_parse": {
        "paper_id": "2005",
        "_pdf_hash": "",
        "abstract": [
            {
                "text": "Statistical machine translation relies heavily on the available training data. In some cases it is necessary to limit the amount of training data that can be created for or actually used by the systems. We introduce weighting schemes which allow us to sort sentences based on the frequency of unseen n-grams. A second approach uses TF-IDF to rank the sentences. After sorting we can select smaller training corpora and we are able to show that systems trained on much less training data achieve a very competitive performance compared to baseline systems using all available training data.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Abstract",
                "sec_num": null
            }
        ],
        "body_text": [
            {
                "text": "The goal of this research was to decrease the amount of training data that is necessary to train a competitive statistical translation system regardless of the actual test data or its domain. \"Competitive\" here means that the system should not produce significantly worse translations compared to a system trained on a significantly larger amount of data. It is important to note that this is not an adaptation approach as we assume that the test data (and its domain) is not known at the time we select the actual training data.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Introduction",
                "sec_num": "1."
            },
            {
                "text": "Statistical machine translation can be described in a formal way as follows: Here t is the target sentence, and s is the source sentence. P(t) is the target language model and P(s|t) is the translation model used in the decoder. Statistical machine translation searches for the best target sentence from the space defined by the target language model and the translation model. Statistical translation models are usually either phrase-or word-based and include most notably IBM1 to IBM4 and HMM ( [1] , [2] , [3] ). Some recent developments focused on online phrase extraction ( [4] , [5] ). All models use available bilingual training data in the source and target languages to estimate their parameters and approximate the translation probabilities. One of the main problems of Statistical Machine Translation (SMT) is the necessity to have large parallel corpora available. This might not be a big issue for major languages, but it certainly is a problem for languages with fewer resources ( [6] , [7] ). To improve the data situation for these languages it is necessary to hire human translators at enormous costs who translate corpora that can later be used to train SMT systems. Our idea focuses on sorting the available source sentences that should be translated by a human translator according to their approximate importance. The importance is estimated using a frequency based and an information retrieval approach.",
                "cite_spans": [
                    {
                        "start": 497,
                        "end": 500,
                        "text": "[1]",
                        "ref_id": "BIBREF0"
                    },
                    {
                        "start": 503,
                        "end": 506,
                        "text": "[2]",
                        "ref_id": "BIBREF1"
                    },
                    {
                        "start": 509,
                        "end": 512,
                        "text": "[3]",
                        "ref_id": "BIBREF2"
                    },
                    {
                        "start": 579,
                        "end": 582,
                        "text": "[4]",
                        "ref_id": "BIBREF3"
                    },
                    {
                        "start": 585,
                        "end": 588,
                        "text": "[5]",
                        "ref_id": "BIBREF4"
                    },
                    {
                        "start": 995,
                        "end": 998,
                        "text": "[6]",
                        "ref_id": "BIBREF5"
                    },
                    {
                        "start": 1001,
                        "end": 1004,
                        "text": "[7]",
                        "ref_id": "BIBREF6"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Introduction",
                "sec_num": "1."
            },
            {
                "text": "There are three inherently different motivations for the goal of limiting the amount of necessary training data for a competitive translation system. We described those motivations and their applications already in the paper [8] .",
                "cite_spans": [
                    {
                        "start": 225,
                        "end": 228,
                        "text": "[8]",
                        "ref_id": "BIBREF7"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Motivation",
                "sec_num": "2."
            },
            {
                "text": "The main problem of portability of SMT systems to new languages is the involved cost to generate parallel bilingual training data as it is necessary to have sentences translated by human translators. An assumption could be that a 1 million word corpus needs to be translated to a new language in order to build a decent SMT system. A human translator could charge in the range of approximately 0.10-0.25 USD per word depending on the involved languages and the difficulty of the text. The translation of a 1 million word corpus would then cost between 100,000 and 250,000 USD. The concept here is to select the most important sentences from the original 1 million word corpus and have only those translated by the human translators. If it would still be possible to get a similar translation performance with a significantly lower translation effort, a considerable amount of money could be saved. This could especially be applied to low density languages with limited resources ( [6] , [7] ).",
                "cite_spans": [
                    {
                        "start": 981,
                        "end": 984,
                        "text": "[6]",
                        "ref_id": "BIBREF5"
                    },
                    {
                        "start": 987,
                        "end": 990,
                        "text": "[7]",
                        "ref_id": "BIBREF6"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Application 1: Reducing Human Translation Cost",
                "sec_num": null
            },
            {
                "text": "Another possible application is the usage of statistical machine translation on portable small devices like PDAs or cell phones. Those devices tend to have a limited amount of memory available which limits the size of the models the device can actually hold and a larger training corpus will usually result in a larger model. The more recent approaches to online phrase extraction for SMT make it necessary to have the corpus available (and in memory) at the time of translation ([4] , [5] ). Given the upper example, a small device might not be able to hold a 1 million word bilingual corpus but e.g. only a corpus with 200,000 words. The question is now which part of the corpus (especially which sentences) should be selected and put on the device to get the best possible translation system.",
                "cite_spans": [
                    {
                        "start": 479,
                        "end": 483,
                        "text": "([4]",
                        "ref_id": null
                    },
                    {
                        "start": 486,
                        "end": 489,
                        "text": "[5]",
                        "ref_id": "BIBREF4"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Application 2: Translation on Small Devices",
                "sec_num": null
            },
            {
                "text": "Even on larger devices that do not have rigid limitations of memory, the approach could be helpful. The complexity of online phrase extraction and standard training algorithms depends mainly on the size of the bilingual training data. Limiting the size of the training data with the same translation performance on these devices would speed up the translations.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Application 3: Standard Translation System",
                "sec_num": null
            },
            {
                "text": "Another problem is that the still widely used 32 bit machines like the Intel Pentium 4 and AMD Athlon XP series can only address up to 4 gigabytes of memory. There are already bilingual corpora in excess of 4 gigabytes available and therefore it is necessary to select the most important sentences from these corpora to be able to hold them in memory. (The last issue will certainly be resolved by the widespread introduction of 64 bit machines which can theoretically address 17 million terabytes of memory.)",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Application 3: Standard Translation System",
                "sec_num": null
            },
            {
                "text": "This research can generally be regarded as an example of active learning. This means the machine learning algorithm does not just passively train on the available training data but plays an active role in selecting the best training data.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Previous Work",
                "sec_num": "3."
            },
            {
                "text": "Active learning, as a standard method in machine learning, has been applied to a variety of problems in natural language processing, for example to parsing ( [9] ) and to automatic speech recognition ( [10] ).",
                "cite_spans": [
                    {
                        "start": 158,
                        "end": 161,
                        "text": "[9]",
                        "ref_id": "BIBREF8"
                    },
                    {
                        "start": 202,
                        "end": 206,
                        "text": "[10]",
                        "ref_id": "BIBREF9"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Previous Work",
                "sec_num": "3."
            },
            {
                "text": "It is important to note the difference between this approach and approaches to Translation Model Adaptation ( [11] ) or simple subsampling techniques that are based on the actual test data. Here we assume that the test data is not known at selection time so the intention is to get the best possible translation system for every possible test data.",
                "cite_spans": [
                    {
                        "start": 110,
                        "end": 114,
                        "text": "[11]",
                        "ref_id": "BIBREF10"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Previous Work",
                "sec_num": "3."
            },
            {
                "text": "Our previous work in this area focused on improving the ngram (type-) coverage by selecting the sentences based on the number of previously unseen n-grams they contain [8] . Section 4.2 will give a short overview over our previous best method.",
                "cite_spans": [
                    {
                        "start": 168,
                        "end": 171,
                        "text": "[8]",
                        "ref_id": "BIBREF7"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Previous Work",
                "sec_num": "3."
            },
            {
                "text": "The sentences are sorted according to the following very simple algorithm.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Algorithm",
                "sec_num": "4.1."
            },
            {
                "text": "For all sentences that are not in the sorted list Calculate weight of sentences Find sentence with highest weight Add sentence with highest weight to sorted list",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Algorithm",
                "sec_num": "4.1."
            },
            {
                "text": "The interesting part is the calculation of the weight of each sentence. The weight of a sentence will generally depend on the previously selected sentences. We present three different schemes to calculate the importance of a sentence. Section 4.2 presents our previous best selection approach and section 4.3 an approach that weights sentences based on the frequency of the unseen n-grams. The method in section 4.4 uses TF-IDF to find sentences that are different from the already seen sentences.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Algorithm",
                "sec_num": "4.1."
            },
            {
                "text": "As stated earlier our previous work in this area focused on optimizing the sorting of the sentences based on the n-gram coverage.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Previous Best Weighting Scheme",
                "sec_num": "4.2."
            },
            {
                "text": "The best results were achieved using the following weighting term:",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Previous Best Weighting Scheme",
                "sec_num": "4.2."
            },
            {
                "text": "sentence n sentence n grams) unseen ( # ) ( est_weight previous_b 2 1 \u2212 = \u2211 =",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Previous Best Weighting Scheme",
                "sec_num": "4.2."
            },
            {
                "text": "This means for each sentence, which had not been sorted yet, the number of unseen uni-and bigrams was calculated and divided by the length of the sentence (in words). This gave significantly better results than the baseline systems where the sentences were not weighted.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Previous Best Weighting Scheme",
                "sec_num": "4.2."
            },
            {
                "text": "The problem with the previous best system is that every unseen unigram gets the same weight. Words that only occur once in the whole training data will be given the same value as higher frequent and probably more important words. The same is certainly true for low-and high-frequency bigrams. This is why we wanted to make sure that our new weighting schemes focus on high-frequency n-grams and put less weight on lower frequency n-grams. This means the goal here is not necessarily to optimize the coverage of the types but of the tokens. We use the frequency of the n-grams in the training data to estimate their importance. The first term just sums over the frequencies for every unseen n-gram to get the sentence weight.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Weighting of Sentences Based on N-gram Frequency",
                "sec_num": "4.3."
            },
            {
                "text": "= \u2212 \uf8fa \uf8fa \uf8fb \uf8f9 \uf8ef \uf8ef \uf8f0 \uf8ee \u2212 = j n n n sentence 1 grams unseen gram) frequency( ) ( j weight",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "\u2211 \u2211",
                "sec_num": null
            },
            {
                "text": "The parameter j here determines the n-grams that are considered and was set to values of 1, 2 and 3 in the experiments. This means an unseen sentence like \"Where is the hotel?\" will have a high weight, especially for data in the tourism domain because we can assume that every n-gram in this sentence is rather frequent. These simple weighting schemes already show improvements over the baseline systems as shown in the later parts of the paper but they have various shortcomings. They do not take the actual translation cost of the sentence into account. (Translators generally charge per word and not per sentence). This leads to the fact that longer sentences tend to get higher weights than shorter sentences, because they will contain more, and possibly higher frequent, unseen n-grams. The focus on token-coverage is certainly very helpful but longer sentences are more difficult for the training of statistical translation models. (When training the translation model IBM1 for example every possible word alignment between sentences is considered.)",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "\u2211 \u2211",
                "sec_num": null
            },
            {
                "text": "To fix these shortcomings we changed the weighting terms to incorporate the actual length of a sentence by dividing the sum of the frequencies of the unseen n-grams by the length of the sentence:",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "\u2211 \u2211",
                "sec_num": null
            },
            {
                "text": "sentence n sentence j n n \u2211 \u2211 = \u2212 \uf8fa \uf8fa \uf8fb \uf8f9 \uf8ef \uf8ef \uf8f0 \uf8ee \u2212 = 1 grams unseen gram) frequency( ) ( j weight",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "\u2211 \u2211",
                "sec_num": null
            },
            {
                "text": "This changes the weight to -informally speaking -\"newly covered tokens in the training data per word to translate\". As noted earlier the algorithms for training translation models in statistical machine translation usually work better (and faster) on shorter sentences. For this reason we also tried to divide by the square of the length of a sentence which prefers even shorter sentences. Overall the weighting terms can be written as:",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "\u2211 \u2211",
                "sec_num": null
            },
            {
                "text": "i j n n sentence n sentence \u2211 \u2211 = \u2212 \uf8fa \uf8fa \uf8fb \uf8f9 \uf8ef \uf8ef \uf8f0 \uf8ee \u2212 = 1 grams unseen gram) frequency( ) ( j i, weight",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "\u2211 \u2211",
                "sec_num": null
            },
            {
                "text": "We introduce the second parameter i here to indicate the exponent of the sentence length (values used in the experiments were 0, 1 and 2). It is certainly possible to use higher values for i and j but the results indicated that higher values would not produce better results.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "\u2211 \u2211",
                "sec_num": null
            },
            {
                "text": "The second approach for the weighting of sentences is based on a different idea and uses an information retrieval method (TF-IDF) to attach a weight to sentences. The similarity between two documents is then defined as the cosine of the angle between the two vectors.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Weighting of sentences based on TF-IDF",
                "sec_num": "4.4."
            },
            {
                "text": "The idea now is to use TF-IDF to find the most different sentence compared to the already selected sentences and give this one the highest importance -this means we just select the sentence with the lowest TF-IDF score (compared to the already selected sentences) next. The first sentence here has to be randomly selected because there is nothing to compare the available sentences against in the first step. The randomly selected sentence could be:",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Sentence weighting with TF-IDF",
                "sec_num": null
            },
            {
                "text": "Where is the hotel?",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "1.",
                "sec_num": null
            },
            {
                "text": "In the next step the TF-IDF score for every still available sentence compared to this sentence is calculated. Sentences that do not have a single common word with this sentence will get the lowest possible TF-IDF score of 0 and one of those will again be selected, for example:",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "1.",
                "sec_num": null
            },
            {
                "text": "1.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "1.",
                "sec_num": null
            },
            {
                "text": "Where is the hotel? 2.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "1.",
                "sec_num": null
            },
            {
                "text": "I had soup for dinner.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "1.",
                "sec_num": null
            },
            {
                "text": "At some point there will be no more sentences left that only contain unseen words so every sentence will get a positive TF-IDF score. The lowest TF-IDF score will then be for sentences that have the fewest number of already seen words and the highest document frequency for these words. A selected sentence in this example could be:",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "1.",
                "sec_num": null
            },
            {
                "text": "Where is the hotel? 2.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "1.",
                "sec_num": null
            },
            {
                "text": "I had soup for dinner.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "1.",
                "sec_num": null
            },
            {
                "text": "This is fine.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "3.",
                "sec_num": null
            },
            {
                "text": "This sentence only shares the word \"is\" with the already sorted sentences. The word \"is\" most likely has a very high document frequency, thus a low IDF score which leads to an overall low score for this particular sentence. A sentence like \"We ate dinner at a restaurant.\" will get a higher score because the shared word \"dinner\" is certainly less frequent than \"is\" and will get a higher IDF score. The TF score in this example would be the same so it can be ignored. In the next iteration the TF score for \"is\" in the sorted sentences will be higher, which in turn lowers the chances to select another sentence with \"is\". This means overall that this weighting scheme will make sure that at the beginning new and unseen words are covered and it will give more weight to higher frequent words later, which is the same behavior as the weighting schemes presented in section 4.3.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "3.",
                "sec_num": null
            },
            {
                "text": "A more information-retrieval centered motivation for the TF-IDF method could be: We always select the sentence with the topic that is \"furthest away\" from the topic(s) of the sentences we already sorted. This will make sure that we cover all possible topics that are in our training data and might come up in the test data.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "3.",
                "sec_num": null
            },
            {
                "text": "TF-IDF can easily be generalized to n-grams by using every ngram as an entry in the document vectors (instead of only using words). We tried this for n-grams up to bigrams and plan on doing experiments with higher n-grams. The following section 5 will give an overview over the experiments that were done using the three presented approaches to sort sentences according to their estimated importance.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Generalizing TF-IDF for N-grams",
                "sec_num": null
            },
            {
                "text": "The full training data for the translation experiments consisted of 123,416 English sentences with 903,525 English words (tokens). This data is part of the BTEC corpus ( [12] ) with relatively simple sentences from the travel domain. The whole training data was also available in Spanish (852,362 words). The testing data which was used to measure the machine translation performance consisted of 500 lines of data from the medical domain. All translations in this task were done translating English to Spanish.",
                "cite_spans": [
                    {
                        "start": 170,
                        "end": 174,
                        "text": "[12]",
                        "ref_id": "BIBREF11"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Test and Training Data",
                "sec_num": "5.1."
            },
            {
                "text": "The applied statistical machine translation system uses an online phrase extraction algorithm based on IBM1 lexicon probabilities ( [3] , [13] ). The language model is a trigram language model with Kneser-Ney-discounting built with the SRI-Toolkit ( [14] ) using only the Spanish part of the training data.",
                "cite_spans": [
                    {
                        "start": 132,
                        "end": 135,
                        "text": "[3]",
                        "ref_id": "BIBREF2"
                    },
                    {
                        "start": 138,
                        "end": 142,
                        "text": "[13]",
                        "ref_id": "BIBREF12"
                    },
                    {
                        "start": 250,
                        "end": 254,
                        "text": "[14]",
                        "ref_id": "BIBREF13"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Machine Translation System",
                "sec_num": "5.2."
            },
            {
                "text": "We applied the standard metrics introduced for machine translation, NIST ( [15] ) and BLEU ( [16] ). For the baseline systems that do not use all available training data we selected sentences based on the original order of the training corpus and trained the smaller systems from this data. The second \"baseline\" systems were trained using the previous best approach presented in section 4.2.",
                "cite_spans": [
                    {
                        "start": 75,
                        "end": 79,
                        "text": "[15]",
                        "ref_id": "BIBREF14"
                    },
                    {
                        "start": 93,
                        "end": 97,
                        "text": "[16]",
                        "ref_id": "BIBREF15"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Machine Translation System",
                "sec_num": "5.2."
            },
            {
                "text": "Translation systems trained on these (smaller) data sets give the scores shown in diagrams 1 and 2. The diagrams clearly illustrate that after a rather steep increase of the scores until the translation of approximately 400,000 words the scores of the baseline increase only slightly until they reach the final score for the system using all available training data. The previous best selection especially benefits at the beginning for a lower number of translated words and hits a NIST score of 4.0 at 170,000 translated words, which is very close to the confidence interval and only about 5% worse than the best overall score. A NIST score of 4.1 is already achieved at 220,000 translated words and 2% worse than the final baseline of 4.19. At 10,000 translated words the previous best system achieves a NIST score of 2.56, compared to a baseline of 2.04.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Baseline and Previous Best Systems",
                "sec_num": "5.3."
            },
            {
                "text": "1 95% confidence intervals ",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Baseline and Previous Best Systems",
                "sec_num": "5.3."
            },
            {
                "text": "Because of the limited space we will only show diagrams for the NIST scores for each experiment. This can be justified as the graphs for the BLEU scores showed basically the same behavior. We did also not include the graph for the previous best system in the diagrams because the new approaches did not always clearly improve over the previous best system and this would have led to even more close-packed diagrams.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Translation Results",
                "sec_num": "5.4."
            },
            {
                "text": "Diagram 3 illustrates the NIST scores for systems where the sentences were sorted according to weight 0,j . If the optimization only uses the frequency sum of previously unseen unigrams to rank sentences, the systems score significantly higher than the baseline for very small amounts of training data. But the steep increase stops very soon and the systems fall slightly below the baseline, recover towards the end, and finish on the same scores. These problems are clearly fixed by incorporating the bi-and trigrams into the optimization process. The scores no longer fall beyond the scores of the baseline systems but stay consistently higher. The systems optimized on uni-and bigrams (weight 0,2 ) are not significantly different from the systems for uni-/bi-and trigrams (weight 0,3 ) but show a very similar performance with slight advantages for the uni-and bigram-systems. Unfortunately both systems do not outperform the previous best method as they reach a NIST score of 4.0 at 230,000 and 240,000 translated words and a score of 4.1 at 300,000 and 320,000 translated words. However all three systems achieve better NIST scores at very small amounts of training data with the same NIST score of 2.72 for 10,000 translated words. ",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Results for term weight 0,j",
                "sec_num": null
            },
            {
                "text": "The difference between the term weight 0,j and weight 1,j is the incorporation of the length of a sentence. The frequency sum of the unseen n-grams is divided by the number of words in the respective sentence to get the weight for the sentence. Diagram 4 illustrates the associated NIST scores. A comparison with Diagram 3 shows that the NIST scores for the sorting of the sentences according to weight 1,j are even better than for the term weight 0,j . We see a very similar behavior for the unigrams and an improvement for the optimizations based on uni-and bigrams and uni-/bi-and trigrams compared to weight 0,j . We also do not see any significant differences between the scores for those two optimizations. The performance is very similar with only slight advantages for the optimization based on uni-and bi-grams (weight 1,2 ). For this system a NIST score of 4.0 was already reached at 140,000 translated words (190,000 for weight 1,3 ) while 4.1 was reached at 300,000 translated words (280,000 for weight 1,3 ). It is again possible to outperform the baseline and previous best systems at 10,000 translated words with a NIST scores of 2.64 (weight 1,1 ) and 2.97 (weight 1,2 and weight 1,3 ).",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Results for term weight 1,j",
                "sec_num": null
            },
            {
                "text": "As explained in section 4.3 we tried to prefer shorter sentences in term weight 2,j by dividing the frequency sum of the unseen n-grams by the square of the number of words in the respective sentence. Diagram 5 illustrates those scores. The scores overall are similar to the earlier diagrams. The term weight 2,2 reaches a NIST score of 4.0 at 180,000 (220,000 for weight 2,3 ) translated words, and a NIST score of 4.1 at 220,000 translated words (270,000 for weight 2,3 ). The systems again outperform the other systems for 10,000 translated words with NIST scores of 3.02 for weight 2,3 and 2.98 for weight 2,2 (weight 2,1 gets a NIST score of only 2.56). ",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Results for term weight 2,j",
                "sec_num": null
            },
            {
                "text": "Diagram 6 shows the scores for the optimization based on TF-IDF for unigrams and uni-/bigrams. In this case the original TF-IDF (based only on unigrams) slightly outperforms the TF-IDF based on uni-and bigrams but both approaches do not show better results than the earlier weighting terms. Table 1 compares the results achieved by the different methods with a special focus on small amounts of data. We give NIST scores for 10,000; 20,000; 50,000 and 100,000 translated words. The last 2 columns show the number of translated words (in thousands) necessary to achieve NIST scores of 4.0 and 4. One might argue that improvements at very small data sizes are not relevant, as the translations will still be very deficient. This might be the case, but there are applications where even a low-quality translation can be helpful ( [17] ). And as we showed in [8] -some translations are surprisingly good, even for very small amounts of training data.",
                "cite_spans": [
                    {
                        "start": 827,
                        "end": 831,
                        "text": "[17]",
                        "ref_id": "BIBREF16"
                    },
                    {
                        "start": 855,
                        "end": 858,
                        "text": "[8]",
                        "ref_id": "BIBREF7"
                    }
                ],
                "ref_spans": [
                    {
                        "start": 291,
                        "end": 298,
                        "text": "Table 1",
                        "ref_id": null
                    }
                ],
                "eq_spans": [],
                "section": "Results for TF-IDF based sorting",
                "sec_num": null
            },
            {
                "text": "The presented weighting schemes could certainly incorporate other features of the original training data. The pure frequency based approach \"tries\" to cover every n gram once and then does not consider it anymore. It might be helpful to have a goal of covering every n-gram a number of times to get better estimates of translation probabilities. The TF-IDF based sorting did not yet show improvements over the earlier approaches. We hope that it will be beneficial to further investigate this idea and maybe combine it with the other methods. Both presented methods give a high weight to function words at the beginning. This is not necessarily desirable so it could be helpful to lower the impact of function words and increase the weight of (high-frequent) content words. Especially the NIST score could benefit from correctly translated content words, as it incorporates the information gain in the score calculation. It might be reasonable for some applications to also consider the target language part of the training data when sorting the sentences. This is certainly not possible if the goal is to limit the effort for human translators and the target sentences are not even available at selection time. It could however be included in the selection of training data for small devices because here the translations will already be available.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Future Work",
                "sec_num": "6."
            },
            {
                "text": "We presented two new weighting schemes to sort training sentences for statistical machine translation according to their importance for the translation performance. The first method mainly tries to improve the token coverage while taking the sentence length into account. We are able to outperform our baseline and our previously best system and see especially nice improvement for very small data sizes. The focus on token coverage is achieved by using the frequency of the previously unseen n-grams as the basis for the sentence weight.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Conclusion",
                "sec_num": "7."
            },
            {
                "text": "We also presented a second idea that bases the sorting of the sentences on the similarity measure TF-IDF, but we did not see improvements over the first method.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Conclusion",
                "sec_num": "7."
            }
        ],
        "back_matter": [],
        "bib_entries": {
            "BIBREF0": {
                "ref_id": "b0",
                "title": "The mathematics of statistical machine translation: Parameter estimation",
                "authors": [
                    {
                        "first": "Peter",
                        "middle": [
                            "E"
                        ],
                        "last": "Brown",
                        "suffix": ""
                    },
                    {
                        "first": "Stephen",
                        "middle": [
                            "A Della"
                        ],
                        "last": "Pietra",
                        "suffix": ""
                    },
                    {
                        "first": "Vincent",
                        "middle": [
                            "J"
                        ],
                        "last": "Della Pietra",
                        "suffix": ""
                    },
                    {
                        "first": "Robert",
                        "middle": [
                            "L"
                        ],
                        "last": "Mercer",
                        "suffix": ""
                    }
                ],
                "year": 1993,
                "venue": "Computational Linguistics",
                "volume": "19",
                "issue": "2",
                "pages": "263--311",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Peter E. Brown, Stephen A. Della Pietra, Vincent J. Della Pietra, and Robert L. Mercer. 1993. The mathematics of statistical machine translation: Parameter estimation. Computational Linguistics, 19(2), pp. 263-311.",
                "links": null
            },
            "BIBREF1": {
                "ref_id": "b1",
                "title": "HMM-based Word Alignment in Statistical Translation",
                "authors": [
                    {
                        "first": "Stephan",
                        "middle": [],
                        "last": "Vogel",
                        "suffix": ""
                    },
                    {
                        "first": "Hermann",
                        "middle": [],
                        "last": "Ney",
                        "suffix": ""
                    },
                    {
                        "first": "Christoph",
                        "middle": [],
                        "last": "Tillmann",
                        "suffix": ""
                    }
                ],
                "year": 1996,
                "venue": "Proceedings of Coling",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Stephan Vogel, Hermann Ney, and Christoph Tillmann, 1996. HMM-based Word Alignment in Statistical Translation. Proceedings of Coling 1996, Copenhagen, Denmark.",
                "links": null
            },
            "BIBREF2": {
                "ref_id": "b2",
                "title": "The CMU Statistical Translation System",
                "authors": [
                    {
                        "first": "Stephan",
                        "middle": [],
                        "last": "Vogel",
                        "suffix": ""
                    },
                    {
                        "first": "Ying",
                        "middle": [],
                        "last": "Zhang",
                        "suffix": ""
                    },
                    {
                        "first": "Alicia",
                        "middle": [],
                        "last": "Tribble",
                        "suffix": ""
                    },
                    {
                        "first": "Fei",
                        "middle": [],
                        "last": "Huang",
                        "suffix": ""
                    },
                    {
                        "first": "Ashish",
                        "middle": [],
                        "last": "Venugopal",
                        "suffix": ""
                    },
                    {
                        "first": "Bing",
                        "middle": [],
                        "last": "Zhao",
                        "suffix": ""
                    },
                    {
                        "first": "Alex",
                        "middle": [],
                        "last": "Waibel",
                        "suffix": ""
                    }
                ],
                "year": 2003,
                "venue": "Proceedings of MT Summit IX",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Stephan Vogel, Ying Zhang, Alicia Tribble, Fei Huang, Ashish Venugopal, Bing Zhao, and Alex Waibel. 2003. The CMU Statistical Translation System. Proceedings of MT Summit IX, 2003. New Orleans, LA, USA.",
                "links": null
            },
            "BIBREF3": {
                "ref_id": "b3",
                "title": "Scaling Phrase-Based Statistical Machine Translation to Larger Corpora and Longer Phrases",
                "authors": [
                    {
                        "first": "Chris",
                        "middle": [],
                        "last": "Callison-Burch",
                        "suffix": ""
                    },
                    {
                        "first": "Colin",
                        "middle": [],
                        "last": "Bannard",
                        "suffix": ""
                    },
                    {
                        "first": "Josh",
                        "middle": [],
                        "last": "Schroeder",
                        "suffix": ""
                    }
                ],
                "year": 2005,
                "venue": "Proceedings of ACL 2005",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Chris Callison-Burch, Colin Bannard and Josh Schroeder. 2005. Scaling Phrase-Based Statistical Machine Translation to Larger Corpora and Longer Phrases. Proceedings of ACL 2005, Ann Arbor, MI, USA.",
                "links": null
            },
            "BIBREF4": {
                "ref_id": "b4",
                "title": "An Efficient Phrase-to-Phrase Alignment Model for Arbitrarily Long Phrases and Large Corpora",
                "authors": [
                    {
                        "first": "Ying",
                        "middle": [],
                        "last": "Zhang",
                        "suffix": ""
                    },
                    {
                        "first": "Stephan",
                        "middle": [],
                        "last": "Vogel",
                        "suffix": ""
                    }
                ],
                "year": 2005,
                "venue": "Proceedings of EAMT 2005",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Ying Zhang and Stephan Vogel. 2005. An Efficient Phrase-to-Phrase Alignment Model for Arbitrarily Long Phrases and Large Corpora. Proceedings of EAMT 2005, Budapest, Hungary.",
                "links": null
            },
            "BIBREF5": {
                "ref_id": "b5",
                "title": "Corpus Resources and Minority Language Engineering",
                "authors": [
                    {
                        "first": "Tony",
                        "middle": [],
                        "last": "Mcenery",
                        "suffix": ""
                    },
                    {
                        "first": "Paul",
                        "middle": [],
                        "last": "Baker",
                        "suffix": ""
                    },
                    {
                        "first": "Lou",
                        "middle": [],
                        "last": "Burnard",
                        "suffix": ""
                    }
                ],
                "year": 2000,
                "venue": "Proceedings of LREC 2000",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Tony McEnery, Paul Baker, Lou Burnard. 2000. Corpus Resources and Minority Language Engineering. Proceedings of LREC 2000, Athens, Greece.",
                "links": null
            },
            "BIBREF6": {
                "ref_id": "b6",
                "title": "A Trainable Transfer-based Machine Translation Approach for Languages with Limited Resources",
                "authors": [
                    {
                        "first": "Alon",
                        "middle": [],
                        "last": "Lavie",
                        "suffix": ""
                    },
                    {
                        "first": "Katharina",
                        "middle": [],
                        "last": "Probst",
                        "suffix": ""
                    },
                    {
                        "first": "Erik",
                        "middle": [],
                        "last": "Peterson",
                        "suffix": ""
                    },
                    {
                        "first": "Stephan",
                        "middle": [],
                        "last": "Vogel",
                        "suffix": ""
                    },
                    {
                        "first": "Lori",
                        "middle": [],
                        "last": "Levin",
                        "suffix": ""
                    }
                ],
                "year": 2004,
                "venue": "Proceedings of EAMT 2004",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Alon Lavie, Katharina Probst, Erik Peterson, Stephan Vogel, Lori Levin, Ariadna Font-Llitj\u00f3s, and Jaime Carbonell. 2004. A Trainable Transfer-based Machine Translation Approach for Languages with Limited Resources. Proceedings of EAMT 2004, Malta.",
                "links": null
            },
            "BIBREF7": {
                "ref_id": "b7",
                "title": "Low Cost Portability for Statistical Machine Translation based on N-gram Coverage",
                "authors": [
                    {
                        "first": "Matthias",
                        "middle": [],
                        "last": "Eck",
                        "suffix": ""
                    },
                    {
                        "first": "Stephan",
                        "middle": [],
                        "last": "Vogel",
                        "suffix": ""
                    },
                    {
                        "first": "Alex",
                        "middle": [],
                        "last": "Waibel",
                        "suffix": ""
                    }
                ],
                "year": 2005,
                "venue": "Proceedings of MTSummit X 2005",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Matthias Eck, Stephan Vogel, and Alex Waibel. 2005. Low Cost Portability for Statistical Machine Translation based on N-gram Coverage. Proceedings of MTSummit X 2005. Phuket, Thailand.",
                "links": null
            },
            "BIBREF8": {
                "ref_id": "b8",
                "title": "Sample selection for statistical parsing",
                "authors": [
                    {
                        "first": "Rebecca",
                        "middle": [],
                        "last": "Hwa",
                        "suffix": ""
                    }
                ],
                "year": 2004,
                "venue": "Computational Linguistics",
                "volume": "30",
                "issue": "3",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Rebecca Hwa. 2004. Sample selection for statistical parsing. Computational Linguistics vol. 30, no. 3.",
                "links": null
            },
            "BIBREF9": {
                "ref_id": "b9",
                "title": "Selective Sampling of Training Data for Speech Recognition",
                "authors": [
                    {
                        "first": "Teresa",
                        "middle": [
                            "M"
                        ],
                        "last": "Kamm",
                        "suffix": ""
                    },
                    {
                        "first": "G",
                        "middle": [
                            "L"
                        ],
                        "last": "Gerard",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Meyer",
                        "suffix": ""
                    }
                ],
                "year": 2002,
                "venue": "Proceedings of HLT 2002",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Teresa. M. Kamm and Gerard G. L. Meyer. 2002. Selective Sampling of Training Data for Speech Recognition. Proceedings of HLT 2002, San Diego, CA, USA.",
                "links": null
            },
            "BIBREF10": {
                "ref_id": "b10",
                "title": "Adaptation of the Translation Model for Statistical Machine Translation based on Information Retrieval",
                "authors": [
                    {
                        "first": "Almut Silja",
                        "middle": [],
                        "last": "Hildebrand",
                        "suffix": ""
                    },
                    {
                        "first": "Matthias",
                        "middle": [],
                        "last": "Eck",
                        "suffix": ""
                    },
                    {
                        "first": "Stephan",
                        "middle": [],
                        "last": "Vogel",
                        "suffix": ""
                    },
                    {
                        "first": "Alex",
                        "middle": [],
                        "last": "Waibel",
                        "suffix": ""
                    }
                ],
                "year": 2005,
                "venue": "Proceedings of EAMT 2005",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Almut Silja Hildebrand, Matthias Eck, Stephan Vogel and Alex Waibel. 2005. Adaptation of the Translation Model for Statistical Machine Translation based on Information Retrieval. Proceedings of EAMT 2005, Budapest, Hungary.",
                "links": null
            },
            "BIBREF11": {
                "ref_id": "b11",
                "title": "Toward a Broad-coverage Bilingual Corpus for Speech Translation of Travel Conversation in the Real World",
                "authors": [
                    {
                        "first": "Toshiyuki",
                        "middle": [],
                        "last": "Takezawa",
                        "suffix": ""
                    },
                    {
                        "first": "Eiichiro",
                        "middle": [],
                        "last": "Sumita",
                        "suffix": ""
                    },
                    {
                        "first": "Fumiaki",
                        "middle": [],
                        "last": "Sugaya",
                        "suffix": ""
                    },
                    {
                        "first": "Hirofumi",
                        "middle": [],
                        "last": "Yamamoto",
                        "suffix": ""
                    },
                    {
                        "first": "Seiichi",
                        "middle": [],
                        "last": "Yamamoto",
                        "suffix": ""
                    }
                ],
                "year": 2002,
                "venue": "Proceedings of LREC 2002",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Toshiyuki Takezawa, Eiichiro Sumita, Fumiaki Sugaya, Hirofumi Yamamoto, and Seiichi Yamamoto. 2002. Toward a Broad-coverage Bilingual Corpus for Speech Translation of Travel Conversation in the Real World. Proceedings of LREC 2002, Las Palmas, Spain.",
                "links": null
            },
            "BIBREF12": {
                "ref_id": "b12",
                "title": "The ISL Statistical Translation System for Spoken Language Translation",
                "authors": [
                    {
                        "first": "Stephan",
                        "middle": [],
                        "last": "Vogel",
                        "suffix": ""
                    },
                    {
                        "first": "Sanjika",
                        "middle": [],
                        "last": "Hewavitharana",
                        "suffix": ""
                    },
                    {
                        "first": "Muntsin",
                        "middle": [],
                        "last": "Kolss",
                        "suffix": ""
                    },
                    {
                        "first": "Alex",
                        "middle": [],
                        "last": "Waibel",
                        "suffix": ""
                    }
                ],
                "year": 2004,
                "venue": "Proceedings of the International Workshop on Spoken Language Translation",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Stephan Vogel, Sanjika Hewavitharana, Muntsin Kolss, and Alex Waibel. 2004. The ISL Statistical Translation System for Spoken Language Translation. Proceedings of the International Workshop on Spoken Language Translation, Kyoto, Japan.",
                "links": null
            },
            "BIBREF13": {
                "ref_id": "b13",
                "title": "SRI Language Modeling Toolkit",
                "authors": [],
                "year": 1995,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "SRI Speech Technology and Research Laboratory. 1995- 2005. SRI Language Modeling Toolkit. http://www.speech.sri.com/projects/srilm/",
                "links": null
            },
            "BIBREF14": {
                "ref_id": "b14",
                "title": "Automatic Evaluation of Machine Translation Quality using n-Gram Cooccurrence Statistics",
                "authors": [
                    {
                        "first": "George",
                        "middle": [],
                        "last": "Doddington",
                        "suffix": ""
                    }
                ],
                "year": 2001,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "George Doddington, 2001. Automatic Evaluation of Machine Translation Quality using n-Gram Co- occurrence Statistics. NIST Washington, DC, USA.",
                "links": null
            },
            "BIBREF15": {
                "ref_id": "b15",
                "title": "BLEU: a Method for Automatic Evaluation of Machine Translation",
                "authors": [
                    {
                        "first": "Kishore",
                        "middle": [],
                        "last": "Papineni",
                        "suffix": ""
                    },
                    {
                        "first": "Salim",
                        "middle": [],
                        "last": "Roukos",
                        "suffix": ""
                    },
                    {
                        "first": "Todd",
                        "middle": [],
                        "last": "Ward",
                        "suffix": ""
                    },
                    {
                        "first": "Wei-Jing",
                        "middle": [],
                        "last": "Zhu",
                        "suffix": ""
                    }
                ],
                "year": 2002,
                "venue": "Proceedings of ACL 2002",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Kishore Papineni, Salim Roukos, Todd Ward, and Wei- Jing Zhu. 2002. BLEU: a Method for Automatic Evaluation of Machine Translation. Proceedings of ACL 2002, Philadelphia, PA, USA.",
                "links": null
            },
            "BIBREF16": {
                "ref_id": "b16",
                "title": "Building a Statistical Machine Translation System from Scratch: How Much Bang Can We Expect for the Buck?",
                "authors": [
                    {
                        "first": "Ulrich",
                        "middle": [],
                        "last": "Germann",
                        "suffix": ""
                    }
                ],
                "year": 2001,
                "venue": "Proceedings of the Data-Driven MT Workshop of ACL",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Ulrich Germann. 2001. Building a Statistical Machine Translation System from Scratch: How Much Bang Can We Expect for the Buck? Proceedings of the Data-Driven MT Workshop of ACL 2001. Toulouse, France.",
                "links": null
            }
        },
        "ref_entries": {
            "FIGREF0": {
                "num": null,
                "uris": null,
                "type_str": "figure",
                "text": "IDF similarity measure TF-IDF is a similarity measure widely used in information retrieval. The main idea of TF-IDF is to represent each document by a vector in the size of the overall vocabulary. Each document D (this will be a sentence or a set of sentences in our case) is then represented as a vector the size of the vocabulary. The entry k w is calculated as: term frequency (TF) of the k-th word in the vocabulary in the document D i.e. the number of occurrences. \u2022 k idf is the inverse document (IDF) frequency of the k-th term,"
            },
            "FIGREF1": {
                "num": null,
                "uris": null,
                "type_str": "figure",
                "text": "NIST scores for sentences sorted according to weight 0,j"
            },
            "FIGREF2": {
                "num": null,
                "uris": null,
                "type_str": "figure",
                "text": "NIST scores for sentences sorted according to weight 1,j"
            },
            "FIGREF3": {
                "num": null,
                "uris": null,
                "type_str": "figure",
                "text": "uni-/bi-/trigram Diagram 5: NIST scores for sentences sorted according to weight 2,j"
            },
            "FIGREF4": {
                "num": null,
                "uris": null,
                "type_str": "figure",
                "text": "NIST scores for sentences sorted according to TF-IDF"
            }
        }
    }
}