File size: 64,249 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 |
{
"paper_id": "2020",
"header": {
"generated_with": "S2ORC 1.0.0",
"date_generated": "2023-01-19T15:01:04.407749Z"
},
"title": "Investigating an approach for low resource language dataset creation, curation and classification: Setswana and Sepedi",
"authors": [
{
"first": "Vukosi",
"middle": [],
"last": "Marivate",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "University of Pretoria",
"location": {}
},
"email": "vukosi.marivate@cs.up.ac.za"
},
{
"first": "Tshephisho",
"middle": [],
"last": "Sefara",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "University of Pretoria",
"location": {}
},
"email": "tsefara@csir.co.za"
},
{
"first": "Vongani",
"middle": [],
"last": "Chabalala",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "University of Pretoria",
"location": {}
},
"email": ""
},
{
"first": "Keamogetswe",
"middle": [],
"last": "Makhaya",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "University of Pretoria",
"location": {}
},
"email": ""
},
{
"first": "Tumisho",
"middle": [],
"last": "Mokgonyane",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "University of Pretoria",
"location": {}
},
"email": ""
},
{
"first": "Rethabile",
"middle": [],
"last": "Mokoena",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "University of Pretoria",
"location": {}
},
"email": ""
},
{
"first": "Abiodun",
"middle": [],
"last": "Modupe",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "University of Pretoria",
"location": {}
},
"email": ""
}
],
"year": "",
"venue": null,
"identifiers": {},
"abstract": "The recent advances in Natural Language Processing have only been a boon for well represented languages, negating research in lesser known global languages. This is in part due to the availability of curated data and research resources. One of the current challenges concerning low-resourced languages are clear guidelines on the collection, curation and preparation of datasets for different use-cases. In this work, we take on the task of creating two datasets that are focused on news headlines (i.e short text) for Setswana and Sepedi and the creation of a news topic classification task from these datasets. In this study, we document our work, propose baselines for classification, and investigate an approach on data augmentation better suited to low-resourced languages in order to improve the performance of the classifiers.",
"pdf_parse": {
"paper_id": "2020",
"_pdf_hash": "",
"abstract": [
{
"text": "The recent advances in Natural Language Processing have only been a boon for well represented languages, negating research in lesser known global languages. This is in part due to the availability of curated data and research resources. One of the current challenges concerning low-resourced languages are clear guidelines on the collection, curation and preparation of datasets for different use-cases. In this work, we take on the task of creating two datasets that are focused on news headlines (i.e short text) for Setswana and Sepedi and the creation of a news topic classification task from these datasets. In this study, we document our work, propose baselines for classification, and investigate an approach on data augmentation better suited to low-resourced languages in order to improve the performance of the classifiers.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Abstract",
"sec_num": null
}
],
"body_text": [
{
"text": "The most pressing issues with regard to low-resource languages are the lack of sufficient language resources, like features related to automation. In this study, we introduce an investigation of a low-resource language that provides automatic formulation and customisation of new capabilities from existing ones. While there are more than six thousand languages spoken globally, the availability of resources among each of those are extraordinarily unbalanced (Nettle, 1998) . For example, if we focus on language resources annotated on the public domain, as of November 2019, AG corpus released about 496, 835 news articles related to the English language from more than 200 sources 1 . Additionally, the Reuters News Dataset (Lewis, 1997) comprise roughly 10, 788 annotated texts from the Reuters financial newswire. Moreover, the New York Times Annotated Corpusholds over 1.8 million articles (Sandhaus, 2008) . Lastly, Google Translate only supports around 100 languages (Johnson et al., 2017) . significant amount of knowledge exists for only a small number of languages, neglecting 17% out of the world's language categories labelled as low-resource, and there are currently no standard annotated tokens in low-resource languages (Strassel and Tracey, 2016) . This in turn, makes it challenging to develop various mechanisms and tools used for Natural Language Processing (NLP). In South Africa, most of the news websites (private and public) are published in English, despite there being 11 official languages (including English). In this paper, we list the premium newspapers by circulation as per the first Quarter of 2019 (Bureau of Circulations, 2019) ( Table 1) . Currently, there is a lack of information surrounding 8 of the 11 official South African languages, with the exception of English, Afrikaans and isiZulu which contain most of the reported datasets. n this work, we aim to provide a general framework for two of the 11 South African languages, to create an annotated linguistic resource for Setswana and Se-1 http://groups.di.unipi.it/\u02dcgulli pedi news headlines. In this study, we applied data sources of the news headlines from the South African Broadcast Corporation (SABC) 2 , their social media streams and a few acoustic news. Unfortunately, at the time of this study, we did not have any direct access to news reports, and hopefully this study can promote collaboration between the national broadcaster and NLP researchers. The rest of the work is organized as follows. Section 2. discusses prior work that has gone into building local corpora in South Africa and how they have been used. Section 3. presents the proposed approach to build a local news corpora and annotating the corpora with categories. From here, we focus on ways to gather data for vectorization and building word embeddings (needing an expanded corpus). We also release and make pre-trained word embeddings for 2 local languages as part of this work (Marivate and Sefara, 2020a) . Section 4. investigate building classification models for the Setswana and Sepedi news and improve those classifiers using a 2 step augmentation approach inspired by work on hierarchical language models (Yu et al., 2019) . Finally, Section 5. concludes and proposes a path forward for this work.",
"cite_spans": [
{
"start": 460,
"end": 474,
"text": "(Nettle, 1998)",
"ref_id": "BIBREF10"
},
{
"start": 727,
"end": 740,
"text": "(Lewis, 1997)",
"ref_id": "BIBREF5"
},
{
"start": 896,
"end": 912,
"text": "(Sandhaus, 2008)",
"ref_id": "BIBREF12"
},
{
"start": 975,
"end": 997,
"text": "(Johnson et al., 2017)",
"ref_id": "BIBREF3"
},
{
"start": 1236,
"end": 1263,
"text": "(Strassel and Tracey, 2016)",
"ref_id": "BIBREF14"
},
{
"start": 2951,
"end": 2979,
"text": "(Marivate and Sefara, 2020a)",
"ref_id": "BIBREF7"
},
{
"start": 3185,
"end": 3202,
"text": "(Yu et al., 2019)",
"ref_id": "BIBREF17"
}
],
"ref_spans": [
{
"start": 1665,
"end": 1673,
"text": "Table 1)",
"ref_id": "TABREF0"
}
],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1."
},
{
"text": "Creating sizeable language resources for low resource languages is important in improving available data for study (Zoph et al., 2016) and cultural preservation. Focusing on the African continent, we note that there are few annotated datasets that are openly available for Natural Language Processing tasks such as classification. In South Africa, the South African Center for Digital Language Resources (SADiLaR) 3 has worked to curate datasets of local South African languages. There remain gaps such as accessing large corpora and data from sources such as broadcasters and news organizations as they have sizeable catalogs that are yet to make it into the public domain. In this work, we work to fill such a gap by collecting, annotating and training classifier models for news headlines in Setswana and Sepedi. As the data that we do find publicly is still small, we also have to deal with the challenges of Machine Learning on small data. Machine learning systems perform poorly in presence of small training sets due to overfitting. To avoid this problem, data augmentation can be used. The technique is well known in the field of image processing (Cubuk et al., 2019) . Data augmentation refers to the augmentation of the training set with artificial, generated, training examples. This technique is used less frequently in NLP but a number of few studies applied data augmentation.",
"cite_spans": [
{
"start": 115,
"end": 134,
"text": "(Zoph et al., 2016)",
"ref_id": "BIBREF19"
},
{
"start": 1155,
"end": 1175,
"text": "(Cubuk et al., 2019)",
"ref_id": "BIBREF2"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Prior Work",
"sec_num": "2."
},
{
"text": "Silfverberg et al. 2017use data augmentation to counteract overfitting where recurrent neural network (RNN) Encoder-Decoder is implemented specifically geared toward a low-resource setting. Authors apply data augmentation by finding words that share word stem for example fizzle and fizzling share fizzl. Then authors replace a stem with another string.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Prior Work",
"sec_num": "2."
},
{
"text": "Zhang et al. (2015) apply data augmentation by using synonyms as substitute words for the original words. However, Kobayashi (2018) states that synonyms are very limited and the synonym-based augmentation cannot produce numerous different patterns from the original texts. Hence, Kobayashi (2018) proposes contextual data augmentation by replacing words that are predicted by a language model given the context surrounding the original words to be augmented.",
"cite_spans": [
{
"start": 115,
"end": 131,
"text": "Kobayashi (2018)",
"ref_id": "BIBREF4"
},
{
"start": 280,
"end": 296,
"text": "Kobayashi (2018)",
"ref_id": "BIBREF4"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Prior Work",
"sec_num": "2."
},
{
"text": "As Wei and Zou (2019) states that these techniques are valid, they are not often used in practice because they have a high cost of implementation relative to performance gain. They propose an easy data augmentation as techniques for boosting performance on text classification tasks. These techniques involve synonym replacement, random insertion, random swap, and random deletion of a word. Authors observed good performance when using fraction of the dataset (%): 1, 5, 10, 20, 30, 40, 50, 60, 70, 80, 90, 100, as the data size increases, the accuracy also increases for augmented and original data. Original data obtained highest accuracy of 88.3% at 100% data size while augmented data obtained accuracy of 88.6% at 50% data size.",
"cite_spans": [
{
"start": 3,
"end": 21,
"text": "Wei and Zou (2019)",
"ref_id": "BIBREF16"
},
{
"start": 466,
"end": 512,
"text": "1, 5, 10, 20, 30, 40, 50, 60, 70, 80, 90, 100,",
"ref_id": null
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Prior Work",
"sec_num": "2."
},
{
"text": "3 www.sadilar.org",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Prior Work",
"sec_num": "2."
},
{
"text": "In this work, we investigate the development of a 2 step text augmentation method in order to be improve classification models for Setswana and Sepedi. To do this we had to first identify a suitable data source. Collect the data, and then annotate the datasets with news categories. After the data is collected and annotated, we then worked to create classification models from the data as is and then use a word embedding and document embedding augmentation approach. In this section discuss how data was collected as well as the approach we use to build classification models.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Developing news classification models for Setswana and Sepedi",
"sec_num": "3."
},
{
"text": "Before we can train classification models, we first have to collect data for 2 distinct processes. First, we present our collected news dataset as well as its annotation. We then discuss how we collected larger datasets for better vectorisation.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Data Collection, Cleaning and Annotation",
"sec_num": "3.1."
},
{
"text": "The news data we collected is from the SABC 4 Facebook pages. The SABC is the public broadcaster for South Africa. Specifically, data was collected from Motsweding FM (An SABC Setswana radio station) 5 and Thobela FM (An SABC Sepedi radio station) 6 . We scraped the news headlines that are published as posts on both Facebook pages. We claim no copyright for the content but used the data for research purposes. We summarize the datasets in Table 2 . We visualize the token distributions in Setswana and Sepedi in Figures 1 and 2 respectively. As can be seen, the datasets are relatively small and as such, we have to look at other ways to build vectorizers that can better generalize as the word token diversity would be very low.",
"cite_spans": [
{
"start": 200,
"end": 201,
"text": "5",
"ref_id": null
}
],
"ref_spans": [
{
"start": 442,
"end": 449,
"text": "Table 2",
"ref_id": "TABREF1"
}
],
"eq_spans": [],
"section": "News data collection and annotation",
"sec_num": "3.1.1."
},
{
"text": "We annotated the datasets by categorizing the news headlines into: Legal, General News,Sports, Other, Politics, Traffic News, Community Activities, Crime, Business and Foreign Affairs. Annotation was done after reading the headlines and coming up with categories that fit both datasets. We show the distribution of the labels in both the Setswana and Sepedi data sets in Figures 3 and 4 (Table 3) . We also make these vectorizers available for other researchers to use. ",
"cite_spans": [],
"ref_spans": [
{
"start": 371,
"end": 386,
"text": "Figures 3 and 4",
"ref_id": "FIGREF2"
},
{
"start": 387,
"end": 396,
"text": "(Table 3)",
"ref_id": "TABREF3"
}
],
"eq_spans": [],
"section": "News data collection and annotation",
"sec_num": "3.1.1."
},
{
"text": "We explore the use of a few classification algorithms to train news classification models. Specifically we train",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "News Classification Models",
"sec_num": "3.2."
},
{
"text": "\u2022 Logistic Regression,",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "News Classification Models",
"sec_num": "3.2."
},
{
"text": "\u2022 Support Vector Classification,",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "News Classification Models",
"sec_num": "3.2."
},
{
"text": "\u2022 XGBoost, and",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "News Classification Models",
"sec_num": "3.2."
},
{
"text": "\u2022 MLP Neural Network.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "News Classification Models",
"sec_num": "3.2."
},
{
"text": "To deal with the challenge of having a small amount of data on short text, we use data augmentation methods, specifically a word embedding based augmentation (Wang and Yang, 2015) , approach that has been shown to work well on short text (Marivate and Sefara, 2019) . We use this approach since we are not able to use other augmentation methods such as synonym based (requires developed Wordnet Synsets (Kobayashi, 2018) ), language models (larger corpora needed train) and back-translation (not readily available for South African languages). We develop and present the use of both word and document embeddings (as an augmentation quality check) inspired by a hierarchical approach to augmentation (Yu et al., 2019) .",
"cite_spans": [
{
"start": 158,
"end": 179,
"text": "(Wang and Yang, 2015)",
"ref_id": "BIBREF15"
},
{
"start": 238,
"end": 265,
"text": "(Marivate and Sefara, 2019)",
"ref_id": "BIBREF6"
},
{
"start": 403,
"end": 420,
"text": "(Kobayashi, 2018)",
"ref_id": "BIBREF4"
},
{
"start": 699,
"end": 716,
"text": "(Yu et al., 2019)",
"ref_id": "BIBREF17"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "News Classification Models",
"sec_num": "3.2."
},
{
"text": "This Section presents the experiments and results. As this is still work in progress, we present some avenues explored in both training classifiers and evaluating them for the task of news headline classification for Setswana and Sepedi.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Experiments and Results",
"sec_num": "4."
},
{
"text": "For each classification problem, we perform 5 fold cross validation. For the bag-of-words and TFIDF vectorizers, we use a maximum token size of 20,000. For word embeddings and language embeddings we use size 50. All vectorizers were trained on the large corpora presented earlier.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Experimental Setup",
"sec_num": "4.1."
},
{
"text": "We run the baseline experiments with the original data using 5-fold cross validation. We show the performance (in terms of weighted F1 score) in the Figures 5 and 6 . We show the baseline results as orig. For both the Bag-of-Words (TF) and TFIDF, the MLP performs very well comparatively to the other methods. In general the TFIDF performs better. ",
"cite_spans": [],
"ref_spans": [
{
"start": 149,
"end": 164,
"text": "Figures 5 and 6",
"ref_id": "FIGREF4"
}
],
"eq_spans": [],
"section": "Baseline Experiments",
"sec_num": "4.1.1."
},
{
"text": "We applied augmentation in different ways. First for Sepedi and Setswana word embeddings (word2vec), we use word embedding-based augmentation. We augment each dataset 20 times on the training data while the validation data is left intact so as to be comparable to the earlier baselines. We show the effect of augmentation in Figures 5 and 6 (performance labeled with aug). The contextual, word2vec based, word augmentation improves the performance of most of the classifiers. If we now introduce a quality check using doc2vec (Algorithm 1) we also notice the impact on the performance for Sepedi (Figure 6 aug qual ) . We were not able to complete experiments with Setswana for the contextual augmentation with a quality check, but will continue working to better under stand the impact of such an algorithm in general. For example, it remains further work to investigate the effects of different similarity thresholds for the algorithm on the overall performance, how such an algorithm works on highly resourced languages vs low resourced languages, how we can make the algorithm efficient etc. It also interesting to look at how performance of classifiers that were only trained with word2vec features would fair. Deep neural networks are not used in this current work and as such we did not use recurrent neural networks, but we can create sentence features from -word2vec by either using: the mean of all word vectors in a sentence, the median of all word vectors in a sentence or the concatenated power means (R\u00fcckl\u00e9 et al., 2018) . We show the performance of using this approach with the classifiers used for Bag of Words and TFIDF earlier in Figure 7 . The performance for this approach is slightly worse with Algorithm 1: Contextual (Word2vec-based) augmentation algorithm with a doc2vec quality check Input: s: a sentence, run: maximum number of attempts at augmentation Output:\u015d a sentence with words replaced 1 def Augment(s,run): the best results for Sepedi news headline classification being with XGBoost on the augmented data. We hope to improve this performance using word2vec feature vectors using recurrent neural networks but currently are of the view that increasing the corpora sizes and the diversity of corpora for the pre-trained word embeddings may yield even better results. Finally, we show the confusion matrix of the best model in Sepedi on a test set in Figure 8 . The classifier categorizes General News, Politics and Legal news headlines best. For others there is more error. A larger news headline dataset is required and classification performance will also need to be compared to models trained on full news data (with the article body). For the Setswana classifiers, the confusion matrix shows that the data skew results in models that mostly can categorize between categories General News and Other. We need to look at re-sampling techniques to improve this performance as well as increasing the initial dataset size.",
"cite_spans": [
{
"start": 1514,
"end": 1535,
"text": "(R\u00fcckl\u00e9 et al., 2018)",
"ref_id": "BIBREF11"
}
],
"ref_spans": [
{
"start": 596,
"end": 616,
"text": "(Figure 6 aug qual )",
"ref_id": "FIGREF5"
},
{
"start": 1649,
"end": 1657,
"text": "Figure 7",
"ref_id": "FIGREF7"
},
{
"start": 2383,
"end": 2391,
"text": "Figure 8",
"ref_id": "FIGREF8"
}
],
"eq_spans": [],
"section": "Augmentation",
"sec_num": "4.1.2."
},
{
"text": "2 Let #\u00bb V",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Augmentation",
"sec_num": "4.1.2."
},
{
"text": "This work introduced the collection and annotation of Setswana and Sepedi news headline data. It remains a challenge that in South Africa, 9 of the 11 official languages have little data such as this that is available to researchers in order to build downstream models that can be used in different applications. Through this work we hope to provide an example of what may be possible even when we have a limited annotated dataset. We exploit the availability of other free text data in Setswana and Sepedi in order to build pre-trained vectorizers for the languages (which are released as part of this work) and then train classification models for news categories. It remains future work to collect more local language news headlines and text to train more models. We have identified other government news sources that can be used. On training embedding models with the data we have collected, further studies are needed to look at how augmentation using the embedding models improve the quality of augmentation.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Conclusion and Future Work",
"sec_num": "5."
},
{
"text": "Bojanowski, P., Grave, E., Joulin, A., and Mikolov, T. (2017) . Enriching word vectors with subword informa-",
"cite_spans": [
{
"start": 39,
"end": 61,
"text": "and Mikolov, T. (2017)",
"ref_id": null
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Bibliographical References",
"sec_num": "6."
},
{
"text": "http://www.sabc.co.za/",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
},
{
"text": "http://www.sabc.co.za/ 5 https://www.facebook.com/MotswedingFM/ 6 https://www.facebook.com/thobelafmyaka/",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
}
],
"back_matter": [],
"bib_entries": {
"BIBREF1": {
"ref_id": "b1",
"title": "Newspaper circulation statistics for the period",
"authors": [
{
"first": "A",
"middle": [],
"last": "Bureau Of Circulations",
"suffix": ""
}
],
"year": 2019,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Bureau of Circulations, A. (2019). Newspaper circulation statistics for the period January-March 2019 (ABC Q1 2019).",
"links": null
},
"BIBREF2": {
"ref_id": "b2",
"title": "Autoaugment: Learning augmentation strategies from data",
"authors": [
{
"first": "E",
"middle": [
"D"
],
"last": "Cubuk",
"suffix": ""
},
{
"first": "B",
"middle": [],
"last": "Zoph",
"suffix": ""
},
{
"first": "D",
"middle": [],
"last": "Mane",
"suffix": ""
},
{
"first": "V",
"middle": [],
"last": "Vasudevan",
"suffix": ""
},
{
"first": "Q",
"middle": [
"V"
],
"last": "Le",
"suffix": ""
}
],
"year": 2019,
"venue": "Proceedings of the IEEE conference on computer vision and pattern recognition",
"volume": "",
"issue": "",
"pages": "113--123",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Cubuk, E. D., Zoph, B., Mane, D., Vasudevan, V., and Le, Q. V. (2019). Autoaugment: Learning augmentation strategies from data. In Proceedings of the IEEE confer- ence on computer vision and pattern recognition, pages 113-123.",
"links": null
},
"BIBREF3": {
"ref_id": "b3",
"title": "Google's multilingual neural machine translation system: Enabling zero-shot translation",
"authors": [
{
"first": "M",
"middle": [],
"last": "Johnson",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Schuster",
"suffix": ""
},
{
"first": "Q",
"middle": [
"V"
],
"last": "Le",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Krikun",
"suffix": ""
},
{
"first": "Y",
"middle": [],
"last": "Wu",
"suffix": ""
},
{
"first": "Z",
"middle": [],
"last": "Chen",
"suffix": ""
},
{
"first": "N",
"middle": [],
"last": "Thorat",
"suffix": ""
},
{
"first": "F",
"middle": [],
"last": "Vi\u00e9gas",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Wattenberg",
"suffix": ""
},
{
"first": "G",
"middle": [],
"last": "Corrado",
"suffix": ""
}
],
"year": 2017,
"venue": "Transactions of the Association for Computational Linguistics",
"volume": "5",
"issue": "",
"pages": "339--351",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Johnson, M., Schuster, M., Le, Q. V., Krikun, M., Wu, Y., Chen, Z., Thorat, N., Vi\u00e9gas, F., Wattenberg, M., Cor- rado, G., et al. (2017). Google's multilingual neural ma- chine translation system: Enabling zero-shot translation. Transactions of the Association for Computational Lin- guistics, 5:339-351.",
"links": null
},
"BIBREF4": {
"ref_id": "b4",
"title": "Contextual augmentation: Data augmentation by words with paradigmatic relations",
"authors": [
{
"first": "S",
"middle": [],
"last": "Kobayashi",
"suffix": ""
}
],
"year": 2018,
"venue": "Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies",
"volume": "2",
"issue": "",
"pages": "452--457",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Kobayashi, S. (2018). Contextual augmentation: Data augmentation by words with paradigmatic relations. In Proceedings of the 2018 Conference of the North Ameri- can Chapter of the Association for Computational Lin- guistics: Human Language Technologies, Volume 2 (Short Papers), volume 2, pages 452-457.",
"links": null
},
"BIBREF5": {
"ref_id": "b5",
"title": "Reuters-21578 text categorization collection data set",
"authors": [
{
"first": "D",
"middle": [
"D"
],
"last": "Lewis",
"suffix": ""
}
],
"year": 1997,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Lewis, D. D. (1997). Reuters-21578 text categorization collection data set.",
"links": null
},
"BIBREF6": {
"ref_id": "b6",
"title": "Improving short text classification through global augmentation methods",
"authors": [
{
"first": "V",
"middle": [],
"last": "Marivate",
"suffix": ""
},
{
"first": "T",
"middle": [],
"last": "Sefara",
"suffix": ""
}
],
"year": 2019,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"arXiv": [
"arXiv:1907.03752"
]
},
"num": null,
"urls": [],
"raw_text": "Marivate, V. and Sefara, T. (2019). Improving short text classification through global augmentation methods. arXiv preprint arXiv:1907.03752.",
"links": null
},
"BIBREF7": {
"ref_id": "b7",
"title": "African embeddings",
"authors": [
{
"first": "V",
"middle": [],
"last": "Marivate",
"suffix": ""
},
{
"first": "T",
"middle": [],
"last": "Sefara",
"suffix": ""
}
],
"year": 2020,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"DOI": [
"10.5281/zenodo.3668481"
]
},
"num": null,
"urls": [],
"raw_text": "Marivate, V. and Sefara, T. (2020a). African em- beddings [nlp]. https://doi.org/10.5281/ zenodo.3668481, February.",
"links": null
},
"BIBREF8": {
"ref_id": "b8",
"title": "South African news data dataset",
"authors": [
{
"first": "V",
"middle": [],
"last": "Marivate",
"suffix": ""
},
{
"first": "T",
"middle": [],
"last": "Sefara",
"suffix": ""
}
],
"year": 2020,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"DOI": [
"10.5281/zenodo.3668489"
]
},
"num": null,
"urls": [],
"raw_text": "Marivate, V. and Sefara, T. (2020b). South African news data dataset. https://doi.org/10.5281/ zenodo.3668489.",
"links": null
},
"BIBREF9": {
"ref_id": "b9",
"title": "Distributed representations of words and phrases and their compositionality",
"authors": [
{
"first": "T",
"middle": [],
"last": "Mikolov",
"suffix": ""
},
{
"first": "I",
"middle": [],
"last": "Sutskever",
"suffix": ""
},
{
"first": "K",
"middle": [],
"last": "Chen",
"suffix": ""
},
{
"first": "G",
"middle": [
"S"
],
"last": "Corrado",
"suffix": ""
},
{
"first": "J",
"middle": [],
"last": "Dean",
"suffix": ""
}
],
"year": 2013,
"venue": "Advances in neural information processing systems",
"volume": "",
"issue": "",
"pages": "3111--3119",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Mikolov, T., Sutskever, I., Chen, K., Corrado, G. S., and Dean, J. (2013). Distributed representations of words and phrases and their compositionality. In Advances in neural information processing systems, pages 3111- 3119.",
"links": null
},
"BIBREF10": {
"ref_id": "b10",
"title": "Explaining global patterns of language diversity",
"authors": [
{
"first": "D",
"middle": [],
"last": "Nettle",
"suffix": ""
}
],
"year": 1998,
"venue": "Journal of anthropological archaeology",
"volume": "17",
"issue": "4",
"pages": "354--374",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Nettle, D. (1998). Explaining global patterns of lan- guage diversity. Journal of anthropological archaeol- ogy, 17(4):354-374.",
"links": null
},
"BIBREF11": {
"ref_id": "b11",
"title": "Concatenated power mean word embeddings as universal cross-lingual sentence representations",
"authors": [
{
"first": "A",
"middle": [],
"last": "R\u00fcckl\u00e9",
"suffix": ""
},
{
"first": "S",
"middle": [],
"last": "Eger",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Peyrard",
"suffix": ""
},
{
"first": "I",
"middle": [],
"last": "Gurevych",
"suffix": ""
}
],
"year": 2018,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"arXiv": [
"arXiv:1803.01400"
]
},
"num": null,
"urls": [],
"raw_text": "R\u00fcckl\u00e9, A., Eger, S., Peyrard, M., and Gurevych, I. (2018). Concatenated power mean word embeddings as universal cross-lingual sentence representations. arXiv preprint arXiv:1803.01400.",
"links": null
},
"BIBREF12": {
"ref_id": "b12",
"title": "The new york times annotated corpus. Linguistic Data Consortium",
"authors": [
{
"first": "E",
"middle": [],
"last": "Sandhaus",
"suffix": ""
}
],
"year": 2008,
"venue": "",
"volume": "6",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Sandhaus, E. (2008). The new york times annotated corpus. Linguistic Data Consortium, Philadelphia, 6(12):e26752.",
"links": null
},
"BIBREF13": {
"ref_id": "b13",
"title": "Data augmentation for morphological reinflection",
"authors": [
{
"first": "M",
"middle": [],
"last": "Silfverberg",
"suffix": ""
},
{
"first": "A",
"middle": [],
"last": "Wiemerslage",
"suffix": ""
},
{
"first": "L",
"middle": [],
"last": "Liu",
"suffix": ""
},
{
"first": "L",
"middle": [
"J"
],
"last": "Mao",
"suffix": ""
}
],
"year": 2017,
"venue": "Proceedings of the CoNLL SIGMORPHON 2017 Shared Task: Universal Morphological Reinflection",
"volume": "",
"issue": "",
"pages": "90--99",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Silfverberg, M., Wiemerslage, A., Liu, L., and Mao, L. J. (2017). Data augmentation for morphological reinflec- tion. In Proceedings of the CoNLL SIGMORPHON 2017 Shared Task: Universal Morphological Reinflec- tion, pages 90-99.",
"links": null
},
"BIBREF14": {
"ref_id": "b14",
"title": "Lorelei language packs: Data, tools, and resources for technology development in low resource languages",
"authors": [
{
"first": "S",
"middle": [],
"last": "Strassel",
"suffix": ""
},
{
"first": "J",
"middle": [],
"last": "Tracey",
"suffix": ""
}
],
"year": 2016,
"venue": "Proceedings of the Tenth International Conference on Language Resources and Evaluation (LREC'16)",
"volume": "",
"issue": "",
"pages": "3273--3280",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Strassel, S. and Tracey, J. (2016). Lorelei language packs: Data, tools, and resources for technology development in low resource languages. In Proceedings of the Tenth International Conference on Language Resources and Evaluation (LREC'16), pages 3273-3280.",
"links": null
},
"BIBREF15": {
"ref_id": "b15",
"title": "That's so annoying!!!: A lexical and frame-semantic embedding based data augmentation approach to automatic categorization of annoying behaviors using# petpeeve tweets",
"authors": [
{
"first": "W",
"middle": [
"Y"
],
"last": "Wang",
"suffix": ""
},
{
"first": "D",
"middle": [],
"last": "Yang",
"suffix": ""
}
],
"year": 2015,
"venue": "Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing",
"volume": "",
"issue": "",
"pages": "2557--2563",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Wang, W. Y. and Yang, D. (2015). That's so annoying!!!: A lexical and frame-semantic embedding based data aug- mentation approach to automatic categorization of an- noying behaviors using# petpeeve tweets. In Proceed- ings of the 2015 Conference on Empirical Methods in Natural Language Processing, pages 2557-2563.",
"links": null
},
"BIBREF16": {
"ref_id": "b16",
"title": "Eda: Easy data augmentation techniques for boosting performance on text classification tasks",
"authors": [
{
"first": "J",
"middle": [],
"last": "Wei",
"suffix": ""
},
{
"first": "K",
"middle": [],
"last": "Zou",
"suffix": ""
}
],
"year": 2019,
"venue": "Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP)",
"volume": "",
"issue": "",
"pages": "6383--6389",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Wei, J. and Zou, K. (2019). Eda: Easy data augmentation techniques for boosting performance on text classifica- tion tasks. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Lan- guage Processing (EMNLP-IJCNLP), pages 6383-6389.",
"links": null
},
"BIBREF17": {
"ref_id": "b17",
"title": "Hierarchical data augmentation and the application in text classification",
"authors": [
{
"first": "S",
"middle": [],
"last": "Yu",
"suffix": ""
},
{
"first": "J",
"middle": [],
"last": "Yang",
"suffix": ""
},
{
"first": "D",
"middle": [],
"last": "Liu",
"suffix": ""
},
{
"first": "R",
"middle": [],
"last": "Li",
"suffix": ""
},
{
"first": "Y",
"middle": [],
"last": "Zhang",
"suffix": ""
},
{
"first": "S",
"middle": [],
"last": "Zhao",
"suffix": ""
}
],
"year": 2019,
"venue": "IEEE Access",
"volume": "7",
"issue": "",
"pages": "185476--185485",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Yu, S., Yang, J., Liu, D., Li, R., Zhang, Y., and Zhao, S. (2019). Hierarchical data augmentation and the ap- plication in text classification. IEEE Access, 7:185476- 185485.",
"links": null
},
"BIBREF18": {
"ref_id": "b18",
"title": "Character-level convolutional networks for text classification",
"authors": [
{
"first": "X",
"middle": [],
"last": "Zhang",
"suffix": ""
},
{
"first": "J",
"middle": [],
"last": "Zhao",
"suffix": ""
},
{
"first": "Y",
"middle": [],
"last": "Lecun",
"suffix": ""
}
],
"year": 2015,
"venue": "Advances in neural information processing systems",
"volume": "",
"issue": "",
"pages": "649--657",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Zhang, X., Zhao, J., and LeCun, Y. (2015). Character-level convolutional networks for text classification. In Ad- vances in neural information processing systems, pages 649-657.",
"links": null
},
"BIBREF19": {
"ref_id": "b19",
"title": "Transfer learning for low-resource neural machine translation",
"authors": [
{
"first": "B",
"middle": [],
"last": "Zoph",
"suffix": ""
},
{
"first": "D",
"middle": [],
"last": "Yuret",
"suffix": ""
},
{
"first": "J",
"middle": [],
"last": "May",
"suffix": ""
},
{
"first": "K",
"middle": [],
"last": "Knight",
"suffix": ""
}
],
"year": 2016,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"arXiv": [
"arXiv:1604.02201"
]
},
"num": null,
"urls": [],
"raw_text": "Zoph, B., Yuret, D., May, J., and Knight, K. (2016). Trans- fer learning for low-resource neural machine translation. arXiv preprint arXiv:1604.02201.",
"links": null
}
},
"ref_entries": {
"FIGREF0": {
"text": "Figure 1: Setswana Wordcloud",
"type_str": "figure",
"uris": null,
"num": null
},
"FIGREF1": {
"text": "Figure 3: Setswana news title category distribution",
"type_str": "figure",
"uris": null,
"num": null
},
"FIGREF2": {
"text": "Sepedi news title category distribution way that we could create Bag of Words, TFIDF, Word2vec(Mikolov et al., 2013) andFastText (Bojanowski et al., 2017) vectorizers",
"type_str": "figure",
"uris": null,
"num": null
},
"FIGREF4": {
"text": "Baseline classification model performance for Setswana news title categorization",
"type_str": "figure",
"uris": null,
"num": null
},
"FIGREF5": {
"text": "Baseline classification model performance for Sepedi news title categorization",
"type_str": "figure",
"uris": null,
"num": null
},
"FIGREF6": {
"text": "similar words of w i ; 6 s 0 \u2190 randomly select a word from #\u00bb w given weights as distance; 7\u015d \u2190replace w i with similar word s 0 ;",
"type_str": "figure",
"uris": null,
"num": null
},
"FIGREF7": {
"text": "Word2Vec feature based performance for news headline classification",
"type_str": "figure",
"uris": null,
"num": null
},
"FIGREF8": {
"text": "Confusion Matrix of News headline classification models",
"type_str": "figure",
"uris": null,
"num": null
},
"TABREF0": {
"html": null,
"num": null,
"type_str": "table",
"text": "Top newspapers in South Africa with their lan-",
"content": "<table><tr><td>guages</td><td/><td/></tr><tr><td>Paper</td><td colspan=\"2\">Language Circulation</td></tr><tr><td>Sunday Times</td><td>English</td><td>260132</td></tr><tr><td>Soccer Laduma</td><td>English</td><td>252041</td></tr><tr><td>Daily Sun</td><td>English</td><td>141187</td></tr><tr><td>Rapport</td><td colspan=\"2\">Afrikaans 113636</td></tr><tr><td>Isolezwe</td><td>isiZulu</td><td>86342</td></tr><tr><td>Sowetan</td><td>English</td><td>70120</td></tr><tr><td>Isolezwe ngeSonto</td><td>isiZulu</td><td>65489</td></tr><tr><td colspan=\"2\">Isolezwe ngoMgqibelo isiZulu</td><td>64676</td></tr><tr><td>Son</td><td colspan=\"2\">Afrikaans 62842</td></tr></table>"
},
"TABREF1": {
"html": null,
"num": null,
"type_str": "table",
"text": "",
"content": "<table><tr><td colspan=\"2\">: News Data Sets</td><td/></tr><tr><td/><td colspan=\"2\">Setswana Sepedi</td></tr><tr><td>Corpus Size (headlines)</td><td>219</td><td>491</td></tr><tr><td colspan=\"2\">Number of Tokens (words) 1561</td><td>3018</td></tr></table>"
},
"TABREF2": {
"html": null,
"num": null,
"type_str": "table",
"text": "TrafficTona ya toka Michael Masutha,ore bahlankedi ba kgoro ya ditirelo tsa tshokollo ya bagolegwa bao ba tateditswego dithieeletsong tsa khomisene ya go nyakisisa mabarebare a go gogwa ga mmuso ka nko,ba swanetse go hlalosa gore ke ka lebaka la eng ba sa swanelwa go fegwa mesomong Legal The full dataset is made available online (Marivate and Sefara, 2020b) for further research use and improvements to the annotation 7 . As previously discussed, we used larger corpora to create language vectorizers for downstream NLP tasks. We discuss this next.",
"content": "<table><tr><td/><td/><td/><td/><td colspan=\"3\">Setswana News Title Categories</td><td/></tr><tr><td/><td>General News</td><td/><td/><td/><td/><td/><td/></tr><tr><td/><td>Legal</td><td/><td/><td/><td/><td/><td/></tr><tr><td/><td>Other</td><td/><td/><td/><td/><td/><td/></tr><tr><td/><td>Politics</td><td/><td/><td/><td/><td/><td/></tr><tr><td>Category</td><td>Crime Foreign Affairs</td><td/><td/><td/><td/><td/><td/></tr><tr><td/><td>Traffic News</td><td/><td/><td/><td/><td/><td/></tr><tr><td/><td>Sports</td><td/><td/><td/><td/><td/><td/></tr><tr><td/><td>Community Activities</td><td/><td/><td/><td/><td/><td/></tr><tr><td/><td>Business</td><td/><td/><td/><td/><td/><td/></tr><tr><td/><td>0</td><td>10</td><td>20</td><td>30</td><td>40 Frequency</td><td>50</td><td>60</td><td>70</td></tr></table>"
},
"TABREF3": {
"html": null,
"num": null,
"type_str": "table",
"text": "",
"content": "<table><tr><td colspan=\"3\">: Vectorizer Corpora Sizes in number of lines (num-</td></tr><tr><td>ber of tokens)</td><td/><td/></tr><tr><td>Source</td><td>Setswana</td><td>Sepedi</td></tr><tr><td>Wikipedia</td><td>478(21924) 8</td><td>300(10190) 9</td></tr><tr><td>JW300 10</td><td>874464(70251)</td><td>618275(53004)</td></tr><tr><td>Bible</td><td>31102(42233)</td><td>29723(38709)</td></tr><tr><td colspan=\"2\">Constitution 11 7077(3940)</td><td>6564(3819)</td></tr><tr><td>SADILAR 12</td><td>33144(61766)</td><td>67036(87838)</td></tr><tr><td>Total</td><td colspan=\"2\">946264(152027) 721977(149355)</td></tr></table>"
}
}
}
} |