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"
}
}
}
} |