File size: 64,533 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 |
{
"paper_id": "2020",
"header": {
"generated_with": "S2ORC 1.0.0",
"date_generated": "2023-01-19T12:14:05.661700Z"
},
"title": "Goals, Challenges and Findings of the VLSP 2020 English-Vietnamese News Translation Shared Task",
"authors": [
{
"first": "Thanh-Le",
"middle": [],
"last": "Ha",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "Karlsruhe Institute of Technology",
"location": {
"country": "Germany"
}
},
"email": "thanh-le.ha@kit.edu"
},
{
"first": "Van-Khanh",
"middle": [],
"last": "Tran",
"suffix": "",
"affiliation": {},
"email": ""
},
{
"first": "Kim-Anh",
"middle": [],
"last": "Nguyen",
"suffix": "",
"affiliation": {},
"email": ""
}
],
"year": "",
"venue": null,
"identifiers": {},
"abstract": "This paper reports the VLSP 2020 English-Vietnamese News Translation shared task, which is one of the six shared tasks organized at the seventh annual workshop on Vietnamese Language and Speech Processing (VLSP 2020). In this task, we provided parallel and monolingual data for training machine translation systems translating English texts into Vietnamese, with the focus of news domain. There were 6 teams participating into the tasks, with 13 submissions in total. We performed both automatic and human evaluations on the submissions and presented the results and our findings at the conference. We hope this would boost the research of Vietnamese machine translation community and start maintaining annual machine translation tasks at VLSP conferences.",
"pdf_parse": {
"paper_id": "2020",
"_pdf_hash": "",
"abstract": [
{
"text": "This paper reports the VLSP 2020 English-Vietnamese News Translation shared task, which is one of the six shared tasks organized at the seventh annual workshop on Vietnamese Language and Speech Processing (VLSP 2020). In this task, we provided parallel and monolingual data for training machine translation systems translating English texts into Vietnamese, with the focus of news domain. There were 6 teams participating into the tasks, with 13 submissions in total. We performed both automatic and human evaluations on the submissions and presented the results and our findings at the conference. We hope this would boost the research of Vietnamese machine translation community and start maintaining annual machine translation tasks at VLSP conferences.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Abstract",
"sec_num": null
}
],
"body_text": [
{
"text": "VLSP stands for Vietnamese Language and Speech Processing Consortium. It is an initiative to establish a community working on speech and text processing for the Vietnamese language. The VLSP 2020 is the seventh annual international workshop and evaluation campaign.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "Machine Translation (MT) is one of the six shared tasks for the VLSP evaluation campaign this year and it is the first time that MT is organized as a VLSP shared task after being a trial task in 2013. As an important research problem of Language and Speech Processing (LSP), MT often attracts interests from the research community. However, research in Vietnamese language-related MT often conducted by several R&D departments from big companies and research labs in large universities. In 2015, the prestigious MT campaign IWSLT (Cettolo et al., 2015) , whose conference was organized in Da Nang, Vietnam, featured English-Vietnamese MT as one of the MT task of that year's campaign and it has been the first and only MT evaluation featuring Vietnamese language to date. We set the following goals when organizing this VLSP 2020 MT evaluation campaign:",
"cite_spans": [
{
"start": 530,
"end": 552,
"text": "(Cettolo et al., 2015)",
"ref_id": "BIBREF1"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "\u2022 Reviving a traditional task in any LSP community and making it to be a recurrent event. Encouraging research for Vietnamese-related MT and engaging researcher into solving interesting problems of MT",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "\u2022 Motivating the contribution of free data and basic LSP tools supporting Vietnameserelated MT research.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "\u2022 Extending practical applications of MT into smart tools and workflows, e.g. developing multilingual education channels, fighting again fake news in any languages and overcoming language barrier in business, tourism, entertainment and international communication.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "Concretely, we have the following contributions while organizing VLSP 2020 English-Vietnamese News Translation task:",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "\u2022 Crawl, collect, compile and release free parallel and monolingual datasets for training and testing English-Vietnamese MT systems 1 .",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "\u2022 Establishing a standard benchmark for research on English-Vietnamese Translation.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "\u2022 Conduct automatic and human evaluations of the participating MT systems.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "This paper is organized as follows. We describe the dataset for training and testing MT systems in Section 2. Section 3 lists the participating teams and summarizes the approaches they employed in their systems. Section 4 presents how we evaluated the translation outputs. We then show the evaluation results in Section 5. Finally, we conclude the the paper by giving our findings and drawing our future plans for the task.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "Although English-Vietnamese is the most popular language pair in the Vietnamese MT community, it is currently considered as \"low-resource\" language pair, where there are only a few public English-Vietnamese parallel corpora with adequate quality for training MT systems. They are Wikipedia articles extracted for the Asian Language Treebank project (Riza et al., 2016) , mixed-domain EVB parallel corpus collected by Ngo et al. (2013) , a multilingual corpus of short and basic sentences from Tatoeba project 2 (Tiedemann, 2012 ) and a COVID-19 multilingual corpus created by ELRC 3 and compiled by Tiedemann (2012) . In total, those corpora contain around 75,000 English-Vietnamese sentence pairs. Besides those high quality datasets, OPUS 4 (Tiedemann, 2012), a website collecting translated texts from the web, compiles and publishes clean versions of movie subtitle datasets extracted from OpenSubtitles 5 (Lison and Tiedemann, 2016) , as well as religious news and bible translations. Although they are large corpora with the number of sentence pairs varying from hundreds thousands to more than three millions, they are unusable without any filtering method since the domain are very narrow (religious) and the quality is not good (movie subtitles).",
"cite_spans": [
{
"start": 349,
"end": 368,
"text": "(Riza et al., 2016)",
"ref_id": "BIBREF10"
},
{
"start": 417,
"end": 434,
"text": "Ngo et al. (2013)",
"ref_id": "BIBREF6"
},
{
"start": 511,
"end": 527,
"text": "(Tiedemann, 2012",
"ref_id": "BIBREF14"
},
{
"start": 599,
"end": 615,
"text": "Tiedemann (2012)",
"ref_id": "BIBREF14"
},
{
"start": 910,
"end": 937,
"text": "(Lison and Tiedemann, 2016)",
"ref_id": "BIBREF5"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Dataset",
"sec_num": "2"
},
{
"text": "Parallel Data. We decide to create more parallel data for English Vietnamese. Our crawling sources are high-quality bilingual or multilingual websites of news and one-speaker educational talks of various topics, mostly technology, entertainment and design (hereby referred as TED-like talks). Because those websites are required to convey the original content in English to other languages (including Vietnamese) and often gone through several review stages before publishing, the quality is assured.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Training Data",
"sec_num": "2.1"
},
{
"text": "First, we extracted some basic conversations from English teaching websites and coupled them to the Tatoeba dataset. For the news domain, we crawled the data from and then applied some simple filtering methods to remove short sentences. Finally, we combined the crawled data with the COVID-19 ELRC data to produce a 20,000sentence-pair parallel corpus.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Training Data",
"sec_num": "2.1"
},
{
"text": "For the TED-like domain, we downloaded TED talks monolingual data of English and Vietnamese from WIT3 6 (Cettolo et al., 2012) , then aligned them based on the sentence ids. Furthermore, we extracted a parallel corpus from the subtitles of the TED-like videos uploaded on Amara 7 -a platform to assist its users to produce captions and subtitles of the videos they uploaded. As the result, more than five hundreds thousands sentence pairs were crawled.",
"cite_spans": [
{
"start": 104,
"end": 126,
"text": "(Cettolo et al., 2012)",
"ref_id": "BIBREF0"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Training Data",
"sec_num": "2.1"
},
{
"text": "Since the quality of the large OpenSubtitle dataset varies in movies, we decided to include it into the training data and let the participants choose how to use it. In the end, we released the following training data in which news is the in-domain data:",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Training Data",
"sec_num": "2.1"
},
{
"text": "Monolingual Data. For this evaluation, we provided target monolingual data which is 2 million Vietnamese sentences, crawled from Vietnamese Team Affiliation Submitted Bluesky Unknown 2 EngineMT (Ngo et al., 2020) UET-ICTU 6 Lab-914 (Le and Nguyen, 2020) HUST 2 NLP-HUST HUST 1 THORLab D-Soft 1 RD-VAIS (Pham et al., 2020) TNU-HUST-VAIS 1 newspapers from various topics. The text has adequate quality to train language models or to conduct back translation. Similar to the parallel data, we let the participants decide how to preprocess the data.",
"cite_spans": [
{
"start": 194,
"end": 212,
"text": "(Ngo et al., 2020)",
"ref_id": "BIBREF8"
},
{
"start": 302,
"end": 321,
"text": "(Pham et al., 2020)",
"ref_id": "BIBREF9"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Training Data",
"sec_num": "2.1"
},
{
"text": "Validation Data. While crawling the news data for training, we also reserved a small part to be validation data. We released a development dataset and a public test dataset at the same time with the training data. The development set contains 1007 English-Vietnamese sentence pairs and the public test set contains 1220 English-Vietnamese sentence pairs. The participants could use one of the validation sets to turn their models' hyperparameters and the other sets for choosing the primary system to be submitted.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Validation and Test Data",
"sec_num": "2.2"
},
{
"text": "Official Test Data. We informed the participants in advance that the in-domain data is News, but we did not reveal the theme is Covid-19 News until the report of the evaluation campaign. In order to avoid cheating and accidentally inclusion of the test data into training or validation data, we manually selected up-to-date English news about Covid-19 from international online newspapers and then asked professional translators to translate them into Vietnamese. The translators need to conform some strict guidelines while translating the official test set, in order to keep it high quality. As the result, the official test set contains 789 sentence pairs. We mixed them with other crawled 2000 sentence pairs and distributed the English part to the participants, asking them to produce the Vietnamese translation using their models.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Validation and Test Data",
"sec_num": "2.2"
},
{
"text": "The organizers received submissions from 6 different teams with the total number of 13 submissions. Table 2 lists the teams. Among them, there are only 3 teams sending their paper describing their approaches and models.",
"cite_spans": [],
"ref_spans": [
{
"start": 100,
"end": 107,
"text": "Table 2",
"ref_id": "TABREF2"
}
],
"eq_spans": [],
"section": "Participants and their Approaches",
"sec_num": "3"
},
{
"text": "All of the three teams submitted neural machine translation systems. And all of them implemented their systems using the state-of-the-art Transformer architecture (Vaswani et al., 2017) . The configurations are different, however. In EngineMT and RD-VAIS systems, the number of layers is 4 and the model size is 512 while in Lab-914 the number of layers is 6 and the model size is 1024.",
"cite_spans": [
{
"start": 163,
"end": 185,
"text": "(Vaswani et al., 2017)",
"ref_id": "BIBREF15"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Architecture",
"sec_num": "3.1"
},
{
"text": "In the preprocessing phase, the teams utilized common techniques on the parallel data. They all removed long sentences, tokenized the words simply by white-spaces and applied some casing treatments. In addition, Lab-914 performed those techniques plus further filtering methods to remove noisy sentences from the Vietnamese monolingual corpus. For casing treatments, Lab-914 simply lower-cased the data, RD-VAIS marked capitalized and upper-cased words by some special tokens before lowercasing and EngineMT applied smart casing using Moses toolkit (Koehn et al., 2007) . All the teams performed subword tokenization using Byte-Pair Encoding algorithm (Sennrich et al., 2016b) implemented in subword-nmt 8 framework with the number of merging operations set at 35,000.",
"cite_spans": [
{
"start": 549,
"end": 569,
"text": "(Koehn et al., 2007)",
"ref_id": "BIBREF3"
},
{
"start": 652,
"end": 676,
"text": "(Sennrich et al., 2016b)",
"ref_id": "BIBREF13"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Preprocessing",
"sec_num": "3.2"
},
{
"text": "All the teams employed Back Translation (Sennrich et al., 2016a) as the sole technique to exploit monolingual data. However, each team had different strategies on how to use the monolingual data. Lab-914 used all the monolingual data provided while EngineMT used much smaller monolingual corpus after filtering out most of them using their proposed data selection techniques. RD-VAIS, on the other hand, built two systems different on the Figure 1 : Appraise's main interface to rank translation outputs size of the monolingual corpus. One contains 1 million sentences and the other contains all 2 million sentences. At the end, they chose the system trained with 1 million sentences back translation based on the performance on the public test set.",
"cite_spans": [],
"ref_spans": [
{
"start": 439,
"end": 447,
"text": "Figure 1",
"ref_id": null
}
],
"eq_spans": [],
"section": "Back Translation",
"sec_num": "3.3"
},
{
"text": "We organized the task in a way that we would expect to see some domain adaptation techniques. Lab-914 did not employ any specific domain adaptation when they used all the provided data in their systems and treated the in-domain news data the same as other data. RD-VAIS, besides the monolingual data which is news, they used only the parallel in-domain data. This might affect badly on their systems since the in-domain data is small and most of their training data come from back translated data. EngineMT is the team who employed several domain adaptation approaches. First they select subsets of data, both monolingual and parallel, which are relevant to the in-domain data with their TF-IDF-based data selection technique. Then they fine-tuned their models on the in-domain data and ensembled all the models they had.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Domain Adaptation",
"sec_num": "3.4"
},
{
"text": "VLSP 2020 is the first Machine Translation Evaluation Campaign for Vietnamese that has both automatic and human evaluation. Furthermore, the human evaluation result is used to rank the teams in the campaign.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Evaluation",
"sec_num": "4"
},
{
"text": "For this campaign, we employed two metrics to evaluate the submissions: BLEU and TER. Since BLEU is the most popular automatic evaluation metric in Machine Translation, it is the main metrics to rank submissions in the automatic evaluation section.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Automatic Evaluation",
"sec_num": "4.1"
},
{
"text": "Five experts which are professional translators and interpreters were invited to conduct the human evaluation for 6 primary systems from 6 teams. Each of them was asked to independently rank the translation outputs of 789 sentences. They were required to follow the evaluation guidelines in which the quality of the translations is rated based on two main criteria: Adequacy and Fluency. Adequacy is rated higher than Fluency, however. We used Appraise 9 (Federmann, 2018) -an opensource web-based MT evaluation framework to assist the experts for the evaluation process. Figure 1 is the main interface that the evaluator can rate the outputs of all submitted systems. For each sentence, the evaluator is shown the English source sentence in a context of three sentences: the previous sentence, the currently considered sentence and the followed sentence. Also the golden translation of those three sentences are displayed as the references. The evaluator needs to rank each system's output from 1 (best) to 6 (worst), and tied ranking is allowed for two or more systems having the same translation quality.",
"cite_spans": [],
"ref_spans": [
{
"start": 455,
"end": 472,
"text": "(Federmann, 2018)",
"ref_id": null
},
{
"start": 572,
"end": 580,
"text": "Figure 1",
"ref_id": null
}
],
"eq_spans": [],
"section": "Human Evaluation",
"sec_num": "4.2"
},
{
"text": "The rankings from 5 experts were converted to pair-wise rankings (number of wins, loses and ties between a pair of two systems). Then they were combined into overall scores using a variant of TrueSkill (Sakaguchi et al., 2014) , a sophisticated algorithms considering not only the average number of wins but also how difficult the task is and the variance of each system's translation quality.",
"cite_spans": [
{
"start": 202,
"end": 226,
"text": "(Sakaguchi et al., 2014)",
"ref_id": "BIBREF11"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Human Evaluation",
"sec_num": "4.2"
},
{
"text": "We evaluated all the submissions, including contrastive systems and informed the participants BLEU and TER of their systems. But only the primary systems are ranked, and by their BLEU scores within statistically significant differences (p \u2264 0.05). The ranking of the teams with corresponding BLEU and TER scores are described in Table 3 .",
"cite_spans": [],
"ref_spans": [
{
"start": 329,
"end": 336,
"text": "Table 3",
"ref_id": "TABREF4"
}
],
"eq_spans": [],
"section": "Automatic Evaluation",
"sec_num": "5.1"
},
{
"text": "Excepts the team THORLab seemed to have some errors in their submission, other teams produced decent outputs. Unsurprisingly, EngineMT led the board with a considerably large margin to the second team RD-VAIS, maybe because of their In some internal test, we realized that other data excepts the OpenSubtitle were high quality and would bring improvements to the systems that use them, even their domain are not news. Lab-914 used a large transformer model and all the provided data, but their BLEU score are not on pair with RD-VAIS which used only the small, in-domain parallel data. We looked into their outputs and their system description as an attempt to explain the possible inconsistency and we discovered that they did not recover casing of their outputs. BLEU is based on the number of overlapping n-grams so that it is more sensitive to upper-cased and capitalized words than TER which is based on the accuracy of individual words. Later, the human evaluation verified our discovery.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Automatic Evaluation",
"sec_num": "5.1"
},
{
"text": "As described in Section 4.2, we gathered the ranking of all the systems from 5 experts and produced a unique score for each systems by using the TrueSkill algorithm with the bootstrap resampling at p-level of p \u2264 0.05. Table 4 lists the final ranking of the teams by human evaluation.",
"cite_spans": [],
"ref_spans": [
{
"start": 219,
"end": 226,
"text": "Table 4",
"ref_id": "TABREF6"
}
],
"eq_spans": [],
"section": "Human Evaluation",
"sec_num": "5.2"
},
{
"text": "While in automatic evaluation, EngineMT is ranked first, here it goes runner-up, after Lab-914. This verifies our assumption about casing recovery. The automatic evaluation metrics do consider casing in their calculation, but the evaluators do not, following their evaluation guidelines.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Human Evaluation",
"sec_num": "5.2"
},
{
"text": "VLSP 2020 English-Vietnamese News Translation task is the first official MT task hosted by VLSP organizers and it is also the first Vietnamese-related MT evaluation campaign featuring both automatic and human evaluation. We hope that it would bring scientific and practical values to the VLSP community as well as our society in dealing with the Covid-19 pandemic and in developing useful AI tools.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Findings and Future Plans",
"sec_num": "6"
},
{
"text": "These are our findings from this VLSP 2020 English-Vietnamese News Translation task:",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Findings and Future Plans",
"sec_num": "6"
},
{
"text": "\u2022 English-Vietnamse MT is still a low-resource task with the lack of large-size, high-quality datasets. Back Translation on news data helps improving the overall translation quality. More data, even with mediocre quality and out-of-domain (e.g. OpenSubtitle), when being used to train large models, also brings significantly gains, especially in the human evaluation.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Findings and Future Plans",
"sec_num": "6"
},
{
"text": "\u2022 News might not be a good domain in case we would like to encourage domain adaptation techniques. Monolingual corpora are often crawled from online newspapers and Back Translation might outperform your finest domain adaptation techniques.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Findings and Future Plans",
"sec_num": "6"
},
{
"text": "\u2022 The approaches and techniques are very common and well-known. There is no interesting research finding from the participants.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Findings and Future Plans",
"sec_num": "6"
},
{
"text": "\u2022 There was no submission considering the linguistic characteristics of Vietnamese language or the differences between two languages: English and Vietnamese.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Findings and Future Plans",
"sec_num": "6"
},
{
"text": "\u2022 There were a few participating teams.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Findings and Future Plans",
"sec_num": "6"
},
{
"text": "We would like to continue hosting MT evaluation tasks in the near future with these plans in mind:",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Findings and Future Plans",
"sec_num": "6"
},
{
"text": "\u2022 More language directions in both wellresource and low-resource conditions",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Findings and Future Plans",
"sec_num": "6"
},
{
"text": "\u2022 More data in the popular MT tasks",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Findings and Future Plans",
"sec_num": "6"
},
{
"text": "\u2022 Consider some useful and interesting domains such as medical, law or technical domains.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Findings and Future Plans",
"sec_num": "6"
},
{
"text": "\u2022 Spread the words to attract more participants working on interesting MT tasks.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Findings and Future Plans",
"sec_num": "6"
},
{
"text": "This task is mainly conducted by the Speech and Language Processing Department, VinGroup Big Data Institute with the help of professional translators and interpreters. We would like to thank Vingroup Innovation Foundation (VINIF) for the financial support and the Association for VLSP for organizing support during the campaign.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Acknowledgment",
"sec_num": "7"
},
{
"text": "The datasets are published at https://github. com/thanhleha-kit/EnViCorpora",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
},
{
"text": "https://tatoeba.org/eng/ 3 https://elrc-share.eu/ 4 https://opus.nlpl.eu/ 5 https://www.opensubtitles.org/en",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
},
{
"text": "https://wit3.fbk.eu/ 7 https://amara.org/en/",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
},
{
"text": "https://github.com/rsennrich/ subword-nmt",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
},
{
"text": "https://github.com/cfedermann/ Appraise",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
}
],
"back_matter": [],
"bib_entries": {
"BIBREF0": {
"ref_id": "b0",
"title": "WIT3: Web Inventory of Transcribed and Translated Talks",
"authors": [
{
"first": "Mauro",
"middle": [],
"last": "Cettolo",
"suffix": ""
},
{
"first": "Christian",
"middle": [],
"last": "Girardi",
"suffix": ""
},
{
"first": "Marcello",
"middle": [],
"last": "Federico",
"suffix": ""
}
],
"year": 2012,
"venue": "Conference of European Association for Machine Translation",
"volume": "",
"issue": "",
"pages": "261--268",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Mauro Cettolo, Christian Girardi, and Marcello Fed- erico. 2012. WIT3: Web Inventory of Transcribed and Translated Talks. In Conference of European As- sociation for Machine Translation, pages 261-268.",
"links": null
},
"BIBREF1": {
"ref_id": "b1",
"title": "The IWSLT 2017 Evaluation Campaign",
"authors": [
{
"first": "Mauro",
"middle": [],
"last": "Cettolo",
"suffix": ""
},
{
"first": "Jan",
"middle": [],
"last": "Niehues",
"suffix": ""
},
{
"first": "Sebastian",
"middle": [],
"last": "St\u00fcker",
"suffix": ""
},
{
"first": "Luisa",
"middle": [],
"last": "Bentivogli",
"suffix": ""
},
{
"first": "Roldano",
"middle": [],
"last": "Cattoni",
"suffix": ""
},
{
"first": "Marcello",
"middle": [],
"last": "Federico",
"suffix": ""
}
],
"year": 2015,
"venue": "International Workshop on Spoken Language Translation (IWSLT'15)",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Mauro Cettolo, Jan Niehues, Sebastian St\u00fcker, Luisa Bentivogli, Roldano Cattoni, and Marcello Federico. 2015. The IWSLT 2017 Evaluation Campaign. In International Workshop on Spoken Language Trans- lation (IWSLT'15), Danang, Vietnam.",
"links": null
},
"BIBREF2": {
"ref_id": "b2",
"title": "Appraise -Evaluation Framework for Machine Translation",
"authors": [
{
"first": "Christian",
"middle": [],
"last": "Federmann",
"suffix": ""
}
],
"year": 2018,
"venue": "Proceedings of the 27th International Conference on Computational Linguistics: System Demonstrations",
"volume": "",
"issue": "",
"pages": "86--88",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Christian Federmann. 2018. Appraise -Evaluation Framework for Machine Translation. In Proceed- ings of the 27th International Conference on Compu- tational Linguistics: System Demonstrations, pages 86-88.",
"links": null
},
"BIBREF3": {
"ref_id": "b3",
"title": "Moses: Open source toolkit for statistical machine translation",
"authors": [
{
"first": "Philipp",
"middle": [],
"last": "Koehn",
"suffix": ""
},
{
"first": "Hieu",
"middle": [],
"last": "Hoang",
"suffix": ""
},
{
"first": "Alexandra",
"middle": [],
"last": "Birch",
"suffix": ""
},
{
"first": "Chris",
"middle": [],
"last": "Callison-Burch",
"suffix": ""
},
{
"first": "Marcello",
"middle": [],
"last": "Federico",
"suffix": ""
},
{
"first": "Nicola",
"middle": [],
"last": "Bertoldi",
"suffix": ""
},
{
"first": "Brooke",
"middle": [],
"last": "Cowan",
"suffix": ""
},
{
"first": "Wade",
"middle": [],
"last": "Shen",
"suffix": ""
},
{
"first": "Christine",
"middle": [],
"last": "Moran",
"suffix": ""
},
{
"first": "Richard",
"middle": [],
"last": "Zens",
"suffix": ""
}
],
"year": 2007,
"venue": "Proceedings of the 45th annual meeting of the association for computational linguistics companion",
"volume": "",
"issue": "",
"pages": "177--180",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Philipp Koehn, Hieu Hoang, Alexandra Birch, Chris Callison-Burch, Marcello Federico, Nicola Bertoldi, Brooke Cowan, Wade Shen, Christine Moran, Richard Zens, et al. 2007. Moses: Open source toolkit for statistical machine translation. In Pro- ceedings of the 45th annual meeting of the associ- ation for computational linguistics companion vol- ume proceedings of the demo and poster sessions, pages 177-180.",
"links": null
},
"BIBREF4": {
"ref_id": "b4",
"title": "Vietnamese-English Translation with Transformer and Back Translation in VLSP 2020 Machine Translation Shared Task",
"authors": [
{
"first": "Cuong",
"middle": [],
"last": "Duc",
"suffix": ""
},
{
"first": "Thi Thu Trang",
"middle": [],
"last": "Le",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Nguyen",
"suffix": ""
}
],
"year": 2020,
"venue": "Proceedings of the Sixth Conference of the Association for Vietnamese Language and Speech Processing",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Duc Cuong Le and Thi Thu Trang Nguyen. 2020. Vietnamese-English Translation with Transformer and Back Translation in VLSP 2020 Machine Trans- lation Shared Task. In Proceedings of the Sixth Con- ference of the Association for Vietnamese Language and Speech Processing (VLSP 2020).",
"links": null
},
"BIBREF5": {
"ref_id": "b5",
"title": "OpenSubti-tles2016: Extracting Large Parallel Corpora from Movie and TV Subtitles",
"authors": [
{
"first": "Pierre",
"middle": [],
"last": "Lison",
"suffix": ""
},
{
"first": "J\u00f6rg",
"middle": [],
"last": "Tiedemann",
"suffix": ""
}
],
"year": 2016,
"venue": "Proceedings of the 10th International Conference on Language Resources and Evaluation (LREC'2016)",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Pierre Lison and J\u00f6rg Tiedemann. 2016. OpenSubti- tles2016: Extracting Large Parallel Corpora from Movie and TV Subtitles. In Proceedings of the 10th International Conference on Language Resources and Evaluation (LREC'2016).",
"links": null
},
"BIBREF6": {
"ref_id": "b6",
"title": "EVBCorpus -A",
"authors": [
{
"first": "Hung",
"middle": [],
"last": "Quoc",
"suffix": ""
},
{
"first": "Werner",
"middle": [],
"last": "Ngo",
"suffix": ""
},
{
"first": "Bartholom\u00e4us",
"middle": [],
"last": "Winiwarter",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Wloka",
"suffix": ""
}
],
"year": 2013,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Quoc Hung Ngo, Werner Winiwarter, and Bartholom\u00e4us Wloka. 2013. EVBCorpus -A",
"links": null
},
"BIBREF7": {
"ref_id": "b7",
"title": "Multi-layer English-Vietnamese Bilingual Corpus for Studying Tasks in Comparative Linguistics",
"authors": [],
"year": null,
"venue": "Proceedings of the 11th Workshop on Asian Language Resources",
"volume": "",
"issue": "",
"pages": "1--9",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Multi-layer English-Vietnamese Bilingual Corpus for Studying Tasks in Comparative Linguistics. In Proceedings of the 11th Workshop on Asian Language Resources, pages 1-9.",
"links": null
},
"BIBREF8": {
"ref_id": "b8",
"title": "The UET-ICTU Submissions to the VLSP 2020 News Translation Task",
"authors": [
{
"first": "Thi-Vinh",
"middle": [],
"last": "Ngo",
"suffix": ""
},
{
"first": "Minh-Thuan",
"middle": [],
"last": "Nguyen",
"suffix": ""
},
{
"first": "Hoang-Minh-Cong",
"middle": [],
"last": "Nguyen",
"suffix": ""
},
{
"first": "Phuong-Thai",
"middle": [],
"last": "Hoang-Quan Nguyen",
"suffix": ""
},
{
"first": "Van-Vinh",
"middle": [],
"last": "Nguyen",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Nguyen",
"suffix": ""
}
],
"year": 2020,
"venue": "Proceedings of the Sixth Conference of the Association for Vietnamese Language and Speech Processing",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Thi-Vinh Ngo, Minh-Thuan Nguyen, Hoang-Minh- Cong Nguyen, Hoang-Quan Nguyen, Phuong-Thai Nguyen, and Van-Vinh Nguyen. 2020. The UET- ICTU Submissions to the VLSP 2020 News Trans- lation Task. In Proceedings of the Sixth Confer- ence of the Association for Vietnamese Language and Speech Processing (VLSP 2020).",
"links": null
},
"BIBREF9": {
"ref_id": "b9",
"title": "A Report on the Neural Machine Translation in VLSP Campaign 2020",
"authors": [
{
"first": "Ngoc",
"middle": [
"Phuong"
],
"last": "Pham",
"suffix": ""
},
{
"first": "Quang",
"middle": [],
"last": "Chung Tran",
"suffix": ""
},
{
"first": "Quang",
"middle": [
"Minh"
],
"last": "Nguyen",
"suffix": ""
},
{
"first": "Hong Quang",
"middle": [],
"last": "Nguyen",
"suffix": ""
}
],
"year": 2020,
"venue": "Proceedings of the Sixth Conference of the Association for Vietnamese Language and Speech Processing",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Ngoc Phuong Pham, Quang Chung Tran, Quang Minh Nguyen, and Hong Quang Nguyen. 2020. A Re- port on the Neural Machine Translation in VLSP Campaign 2020. In Proceedings of the Sixth Con- ference of the Association for Vietnamese Language and Speech Processing (VLSP 2020).",
"links": null
},
"BIBREF10": {
"ref_id": "b10",
"title": "Conference of The Oriental Chapter of International Committee for Coordination and Standardization of Speech Databases and Assessment Techniques (O-COCOSDA)",
"authors": [
{
"first": "Hammam",
"middle": [],
"last": "Riza",
"suffix": ""
},
{
"first": "Michael",
"middle": [],
"last": "Purwoadi",
"suffix": ""
},
{
"first": "Teduh",
"middle": [],
"last": "Uliniansyah",
"suffix": ""
},
{
"first": "Ai",
"middle": [],
"last": "Aw",
"suffix": ""
},
{
"first": "Sharifah",
"middle": [],
"last": "Ti",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Mahani Aljunied",
"suffix": ""
},
{
"first": "Chi",
"middle": [],
"last": "Luong",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Mai",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Vu Tat Thang",
"suffix": ""
},
{
"first": "Phuong",
"middle": [],
"last": "Nguyen",
"suffix": ""
},
{
"first": "Vichet",
"middle": [],
"last": "Thai",
"suffix": ""
},
{
"first": "Sethserey",
"middle": [],
"last": "Chea",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Sam",
"suffix": ""
}
],
"year": 2016,
"venue": "",
"volume": "",
"issue": "",
"pages": "1--6",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Hammam Riza, Michael Purwoadi, Teduh Uliniansyah, Aw Ai Ti, Sharifah Mahani Aljunied, Luong Chi Mai, Vu Tat Thang, Nguyen Phuong Thai, Vichet Chea, Sethserey Sam, et al. 2016. Introduction of the Asian Language Treebank. In 2016 Con- ference of The Oriental Chapter of International Committee for Coordination and Standardization of Speech Databases and Assessment Techniques (O- COCOSDA), pages 1-6. IEEE.",
"links": null
},
"BIBREF11": {
"ref_id": "b11",
"title": "Efficient elicitation of annotations for human evaluation of machine translation",
"authors": [
{
"first": "Keisuke",
"middle": [],
"last": "Sakaguchi",
"suffix": ""
},
{
"first": "Matt",
"middle": [],
"last": "Post",
"suffix": ""
},
{
"first": "Benjamin",
"middle": [],
"last": "Van Durme",
"suffix": ""
}
],
"year": 2014,
"venue": "Proceedings of the Ninth Workshop on Statistical Machine Translation",
"volume": "",
"issue": "",
"pages": "1--11",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Keisuke Sakaguchi, Matt Post, and Benjamin Van Durme. 2014. Efficient elicitation of annota- tions for human evaluation of machine translation. In Proceedings of the Ninth Workshop on Statistical Machine Translation, pages 1-11.",
"links": null
},
"BIBREF12": {
"ref_id": "b12",
"title": "Improving Neural Machine Translation Models with Monolingual Data",
"authors": [
{
"first": "Rico",
"middle": [],
"last": "Sennrich",
"suffix": ""
},
{
"first": "Barry",
"middle": [],
"last": "Haddow",
"suffix": ""
},
{
"first": "Alexandra",
"middle": [],
"last": "Birch",
"suffix": ""
}
],
"year": 2016,
"venue": "Association for Computational Linguistics (ACL 2016)",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Rico Sennrich, Barry Haddow, and Alexandra Birch. 2016a. Improving Neural Machine Translation Models with Monolingual Data. In Association for Computational Linguistics (ACL 2016), Berlin, Ger- many.",
"links": null
},
"BIBREF13": {
"ref_id": "b13",
"title": "Neural Machine Translation of Rare Words with Subword Units",
"authors": [
{
"first": "Rico",
"middle": [],
"last": "Sennrich",
"suffix": ""
},
{
"first": "Barry",
"middle": [],
"last": "Haddow",
"suffix": ""
},
{
"first": "Alexandra",
"middle": [],
"last": "Birch",
"suffix": ""
}
],
"year": 2016,
"venue": "Association for Computational Linguistics (ACL 2016)",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Rico Sennrich, Barry Haddow, and Alexandra Birch. 2016b. Neural Machine Translation of Rare Words with Subword Units. In Association for Computa- tional Linguistics (ACL 2016), Berlin, Germany.",
"links": null
},
"BIBREF14": {
"ref_id": "b14",
"title": "Parallel Data, Tools and Interfaces in OPUS",
"authors": [
{
"first": "J\u00f6rg",
"middle": [],
"last": "Tiedemann",
"suffix": ""
}
],
"year": 2012,
"venue": "Proceedings of the 8th International Conference on Language Resources and Evaluation (LREC'2012)",
"volume": "",
"issue": "",
"pages": "2214--2218",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "J\u00f6rg Tiedemann. 2012. Parallel Data, Tools and Inter- faces in OPUS. In Proceedings of the 8th Interna- tional Conference on Language Resources and Eval- uation (LREC'2012), pages 2214-2218.",
"links": null
},
"BIBREF15": {
"ref_id": "b15",
"title": "Attention is all you need",
"authors": [
{
"first": "Ashish",
"middle": [],
"last": "Vaswani",
"suffix": ""
},
{
"first": "Noam",
"middle": [],
"last": "Shazeer",
"suffix": ""
},
{
"first": "Niki",
"middle": [],
"last": "Parmar",
"suffix": ""
},
{
"first": "Jakob",
"middle": [],
"last": "Uszkoreit",
"suffix": ""
},
{
"first": "Llion",
"middle": [],
"last": "Jones",
"suffix": ""
},
{
"first": "Aidan",
"middle": [
"N"
],
"last": "Gomez",
"suffix": ""
},
{
"first": "Lukasz",
"middle": [],
"last": "Kaiser",
"suffix": ""
},
{
"first": "Illia",
"middle": [],
"last": "Polosukhin",
"suffix": ""
}
],
"year": 2017,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"arXiv": [
"arXiv:1706.03762"
]
},
"num": null,
"urls": [],
"raw_text": "Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Lukasz Kaiser, and Illia Polosukhin. 2017. Attention is all you need. arXiv preprint arXiv:1706.03762.",
"links": null
}
},
"ref_entries": {
"TABREF1": {
"content": "<table/>",
"num": null,
"type_str": "table",
"text": "Training Datasets for VLSP 2020 MT task",
"html": null
},
"TABREF2": {
"content": "<table/>",
"num": null,
"type_str": "table",
"text": "The teams participated to VLSP 2020 MT task",
"html": null
},
"TABREF4": {
"content": "<table/>",
"num": null,
"type_str": "table",
"text": "Automatic evaluation results of the MT task",
"html": null
},
"TABREF6": {
"content": "<table><tr><td>domain adaptation techniques. Bluesky and Lab-</td></tr><tr><td>914 were shared the third rank when the differ-</td></tr><tr><td>ences between their BLEU scores is not signifi-</td></tr><tr><td>cantly obvious. Notably, based on TER, Lab-914</td></tr><tr><td>were ranked second, only after EngineMT.</td></tr></table>",
"num": null,
"type_str": "table",
"text": "Human evaluation results of the MT task",
"html": null
}
}
}
} |