File size: 68,488 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 |
{
"paper_id": "O08-4002",
"header": {
"generated_with": "S2ORC 1.0.0",
"date_generated": "2023-01-19T08:02:40.060474Z"
},
"title": "A Study on Consistency Checking Method of Part-Of-Speech Tagging for Chinese Corpora 1",
"authors": [
{
"first": "Hu",
"middle": [],
"last": "Zhang",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "Shanxi University",
"location": {
"postCode": "030006",
"settlement": "Taiyuan",
"country": "China"
}
},
"email": "zhanghu@sxu.edu.cn"
},
{
"first": "Jiaheng",
"middle": [],
"last": "Zheng",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "Shanxi University",
"location": {
"postCode": "030006",
"settlement": "Taiyuan",
"country": "China"
}
},
"email": "jhzheng@sxu.edu.cn"
}
],
"year": "",
"venue": null,
"identifiers": {},
"abstract": "Ensuring consistency of Part-Of-Speech (POS) tagging plays an important role in the construction of high-quality Chinese corpora. After having analyzed the POS tagging of multi-category words in large-scale corpora, we propose a novel classification-based consistency checking method of POS tagging in this paper. Our method builds a vector model of the context of multi-category words along with using the k-NN algorithm to classify context vectors constructed from POS tagging sequences and to judge their consistency. These methods are evaluated on our 1.5M-word corpus. The experimental results indicate that the proposed method is feasible and effective.",
"pdf_parse": {
"paper_id": "O08-4002",
"_pdf_hash": "",
"abstract": [
{
"text": "Ensuring consistency of Part-Of-Speech (POS) tagging plays an important role in the construction of high-quality Chinese corpora. After having analyzed the POS tagging of multi-category words in large-scale corpora, we propose a novel classification-based consistency checking method of POS tagging in this paper. Our method builds a vector model of the context of multi-category words along with using the k-NN algorithm to classify context vectors constructed from POS tagging sequences and to judge their consistency. These methods are evaluated on our 1.5M-word corpus. The experimental results indicate that the proposed method is feasible and effective.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Abstract",
"sec_num": null
}
],
"body_text": [
{
"text": "The construction of high-quality and large-scale corpora has always been a fundamental research area in the field of Chinese natural language processing. In recent years, rapid developments in the fields of machine translation (MT), information retrieval (IR), etc. are demanding more Chinese corpora of higher quality and larger scale. Ensuring the consistency of Part-of-Speech (POS) tagging plays an important role in the construction of high-quality Chinese corpora. In particular, we focus on consistency checking of the POS tagging of multi-tagged words, which consist of same Chinese characters and are nearly synonymous, yet have different grammatical functions. No matter how many different POS tags a multi-category word may be tagged with, it must be assigned the same POS tag when it appears in a similar context.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1."
},
{
"text": "The general process of tagging text in the Chinese corpora consists of two steps: (1) automatic POS tagging is used to generate preliminary POS tags; (2) the preliminary POS tags are checked manually by domain experts. Novel approaches and techniques have been proposed for automatic rule-based and statistics-based POS tagging, and the \"state-of-the-art\" approaches achieve a tagging precision of at least 95% [Zhang et al. 1998; Huang et al. 2004; Xue et al. 2002; Ng et al. 2004] . A vast proportion of the words appearing in Chinese corpora are multi-category words. We have studied the textual data from the 2M-word Chinese corpus published by Peking University, and statistics show that the number of multi-category words is 5,473, covering 11% of the word types, while the number of word tokens is as high as 438,624, the percentage of which is 47%. Relevant works indicate that the different POS tag sets have different Stat. Huang et al. [2004] provided some important facts based on the 5M-word Sinica Corpus, wherein categorically ambiguous words only take up only 4.298% of all lexical items; however, categorically ambiguous words compose over 54.5% of all tokens. When checking the POS tags, human experts may have disagreements or may make mistakes in some cases. After analyzing 10,000 sentences containing the multi-category words, which were extracted from the 2M-word Chinese corpus of Peking University, the number of incorrect tags for multi-category words was found to be 133, which accounts for around 1.33% of the total.",
"cite_spans": [
{
"start": 411,
"end": 430,
"text": "[Zhang et al. 1998;",
"ref_id": "BIBREF12"
},
{
"start": 431,
"end": 449,
"text": "Huang et al. 2004;",
"ref_id": "BIBREF1"
},
{
"start": 450,
"end": 466,
"text": "Xue et al. 2002;",
"ref_id": "BIBREF8"
},
{
"start": 467,
"end": 482,
"text": "Ng et al. 2004]",
"ref_id": "BIBREF2"
},
{
"start": 934,
"end": 953,
"text": "Huang et al. [2004]",
"ref_id": "BIBREF1"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1."
},
{
"text": "So far in the field of POS tagging, most of the works have focused on novel algorithms or techniques for POS tagging. There are only a limited number of studies that have focused on consistency checking of POS tagging. Xing et al. [1999] analyzed the inconsistency phenomena of word segmentation (WS) and POS tagging. Qu and Chen [2003] improved the corpus quality by obtaining POS tagging knowledge from processed corpora. Qian and Zheng [2003] introduced a rule-based consistency checking method that obtained POS tagging knowledge automatically from processed corpora by machine learning (ML) and rough set (RS) methods. For real corpora, Du and Zheng [2001] proposed a rule-based consistency checking method to identify the inconsistency phenomena of POS tagging. However, the algorithms and techniques for automatic consistency checking of POS tagging proposed in the prior researches still have some insufficiencies. For example, the assignment of POS tags of the inconsistent POS tagging that are not included in the instance set needs to be conducted manually.",
"cite_spans": [
{
"start": 219,
"end": 237,
"text": "Xing et al. [1999]",
"ref_id": "BIBREF7"
},
{
"start": 318,
"end": 336,
"text": "Qu and Chen [2003]",
"ref_id": "BIBREF6"
},
{
"start": 424,
"end": 445,
"text": "Qian and Zheng [2003]",
"ref_id": "BIBREF4"
},
{
"start": 642,
"end": 661,
"text": "Du and Zheng [2001]",
"ref_id": "BIBREF0"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1."
},
{
"text": "In this paper, we propose a novel classification-based method to check the consistency of POS tagging. Compared to Zhang et al. [2004] , the proposed method fully considers the mutual relationships of the POS in POS tagging sequence, and adopts transition probability and emission probability to describe the mutual dependencies and the k-NN algorithm to weight the similarity. We evaluated our proposed algorithm on our 1.5M-word corpus. In an open test, our method achieved a precision rate of 85.24% and a recall rate of 85.84%.",
"cite_spans": [
{
"start": 115,
"end": 134,
"text": "Zhang et al. [2004]",
"ref_id": "BIBREF9"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1."
},
{
"text": "The rest of the paper is organized as follows. Section 2 introduces the context vector model of POS tagging sequences. Section 3 describes the proposed classification-based consistency checking algorithm. Section 4 discusses the experimental results. Finally, the concluding remarks are given in Section 5.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "A Study on Consistency Checking Method of 159 Part-Of-Speech Tagging for Chinese Corpora",
"sec_num": null
},
{
"text": "The basic idea of our approach is to use the contextual information of multi-category words to judge whether they are tagged consistently or not. In other words, if a multi-category word appears in two locations and the surrounding words in those two locations are tagged similarly, the multi-category word should be assigned with the same POS tag in those two locations as well. Hence, our approach is based on the context of multi-category words, and we model the context by looking at a window around a given multi-category word and the tagging sequence of this window. In the rest of this section, we describe the vector representation of the context of multi-category words and how to determine various parameters in our vector representations.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Describing the Context of Multi-Category Words",
"sec_num": "2."
},
{
"text": "Our vector representation of context consists of three key components: the POS tags of each word in a context window (POS attribute), the importance of each word to the center multi-category word based on distance (position attribute), and the dependency of POS tags of the center multi-category word and its surrounding words (Dependency Attribute).",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Vector Representation of the Context of Multi-Category Words",
"sec_num": "2.1"
},
{
"text": "Given a multi-category word and its context window of size l, we represent the words in sequential order as (w 1 ,w 2 ,\u2026,w l ) and the POS tags of each word as (t ,t 2 , \u2026,t l ). We also refer to the latter vector as POS tagging sequence.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Vector Representation of the Context of Multi-Category Words",
"sec_num": "2.1"
},
{
"text": "In practice, we choose a proper value of l so that the context window contains a sufficient number of words and the complexity of our algorithm remains relatively low. We will discuss this matter in detail later. In this study, we set the value of l to be 7, which means w 4 is our multi-category word of interest and the context window includes 3 preceding and following words, where w 4-i is the ith preceding word and w 4+i is the ith following word.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Vector Representation of the Context of Multi-Category Words",
"sec_num": "2.1"
},
{
"text": "The POS tagging sequence contains information of the POS of each preceding and following word in a POS tagging sequence as well as the position of each POS tag. The POS of surrounding words may have different effect on determining the POS of the multi-category word, which we refer to as POS attribute and represent it using a matrix as follows.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "POS Attribute",
"sec_num": null
},
{
"text": ". Given a multi-category word with a context window of size l (w 1 ,w 2 ,\u2026,w l ) and its POS tagging sequence (t ,t 2 , \u2026,t l ), the POS attribute matrix Y is an l by m matrix, where the rows indicate the POS tags of the preceding words, the multi-category word, and the following words in the context window, while the columns present tags in the tag set. Y i,j =1 iff the POS tag of t i is c j .",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Definition 1. Suppose we have a tag set of size m(c 1 ,c 2 ,\u2026,c m )",
"sec_num": null
},
{
"text": "For example, consider the POS attribute matrix of \"\u6bd4\u8f83\" in the following sentence:",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Definition 1. Suppose we have a tag set of size m(c 1 ,c 2 ,\u2026,c m )",
"sec_num": null
},
{
"text": "\u6211\u4eec/r \u5c31/d \u80fd/v \u6bd4\u8f83/d \u51c6\u786e/a \u5730/u \u9884\u6d4b/v \u51fa/v \u961f\u5458/n \u7684/u \u6210\u7ee9/n",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Definition 1. Suppose we have a tag set of size m(c 1 ,c 2 ,\u2026,c m )",
"sec_num": null
},
{
"text": "As we let l=7, we look at the word \"\u6bd4\u8f83\" and its 3 preceding and following words. Hence, the POS tagging sequence is (r,d,v,d,a,u,v) . In our study, we used a standard tag set that consists of 25 tags. Suppose the tag set is (n, v, a, d, u, p, r, m, q, c, w, I, f, s, t, b, z, e, o, l, j, h, k, g, y) , then the POS attribute matrix of \"\u6bd4\u8f83\" in this example is:",
"cite_spans": [
{
"start": 116,
"end": 131,
"text": "(r,d,v,d,a,u,v)",
"ref_id": null
},
{
"start": 224,
"end": 299,
"text": "(n, v, a, d, u, p, r, m, q, c, w, I, f, s, t, b, z, e, o, l, j, h, k, g, y)",
"ref_id": null
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Definition 1. Suppose we have a tag set of size m(c 1 ,c 2 ,\u2026,c m )",
"sec_num": null
},
{
"text": "0, 0, 0, 0, 0, 0,1 0, 0, 0,1, 0, 0, 0 0,1, 0, 0, 0, 0, 0 0, 0, 0,1, 0, 0, 0 0, 0,1, 0, 0, 0, 0 0, 0, 0, 0,1, 0, 0 0,1, 0, 0, 0, 0, 0 \u2026\u2026 \u239b \u239e \u239c \u239f \u2026\u2026 \u239c \u239f \u239c \u239f \u2026\u2026 \u239c \u239f = \u2026 \u2026 \u239c \u239f \u239c \u239f \u2026\u2026 \u239c \u239f \u2026\u2026 \u239c \u239f \u239c \u239f \u2026\u2026 \u239d \u23a0 Y",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Definition 1. Suppose we have a tag set of size m(c 1 ,c 2 ,\u2026,c m )",
"sec_num": null
},
{
"text": "Due to the different distances from the multi-category word, the POS of the word before or after the multi-category word may, in a POS tagging sequence, have a different influence on the POS tagging of the multi-category word, which we refer to as position attribute.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Position Attribute",
"sec_num": null
},
{
"text": "Definition2. Given a multi-category word with a context window of size l, suppose the number of preceding or following words is n (i.e., l=2n+1), the position attribute vector V x of the multi-category word is given by",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Position Attribute",
"sec_num": null
},
{
"text": "V x =(d 1 ,\u2026d n ,d n+1 ,\u2026d l )",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Position Attribute",
"sec_num": null
},
{
"text": "where d n+1 is the value of the position attribute of the multi-category word and d n+1-",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Position Attribute",
"sec_num": null
},
{
"text": "i (d n+1+i )",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Position Attribute",
"sec_num": null
},
{
"text": "is the value of the position attribute of the ith preceding (following) word. We further require that d n+1-i = d n+1+i and",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Position Attribute",
"sec_num": null
},
{
"text": "1 1 1 1 ( ) 1 n n n i n i i d d d + +\u2212 ++ = + + = \u2211 .",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Position Attribute",
"sec_num": null
},
{
"text": "We choose a proper position attribute vector so that the multi-category word itself has the highest weight, and the closer the surrounding word, the higher its weight is. If we consider a context window of size 7, based on our preliminary experiments, we chose the following position attribute values: d 1 = d 7 = 1/22;d 2 = d 6 = 1/11;d 3 = d 5 = 2/11; and d 4 = 4/11. Hence, the final position attribute vector used in our study can be written as follows: ",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Position Attribute",
"sec_num": null
},
{
"text": "((1/ 22),(1/11),(2/11),(4/11),(2/11),(1/11), 1/ 22)) X V = ( .",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Position Attribute",
"sec_num": null
},
{
"text": "X V = (",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Position Attribute",
"sec_num": null
},
{
"text": "Note that, if the POS tag in the POS tagging sequence is incorrect, the position attribute value of the corresponding position should be turned into a negative value, so that when the incorrect POS tag appears in a POS tagging sequence, this attribute can correctly show that the incorrect POS tag has had a negative effect on generating the final context vector.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Position Attribute",
"sec_num": null
},
{
"text": "The last attribute we focus on is dependency attribute, which corresponds to the fact that there are mutual dependencies on the appearance of every POS in POS tagging sequences. In particular, we use transition probability and emission probability in Hidden Markov Model (HMM) to capture this dependency.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Dependency Attribute",
"sec_num": null
},
{
"text": "( , ) , ( ) ( , ) i j i f c c T i j i j f c T P c c = =",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Definition 3. Given a tag set of size m(c 1 ,c 2 ,\u2026,c m ), the transition probability table T is an m by m matrix and given by:",
"sec_num": null
},
{
"text": "where f(c i , c j ) is the frequency of the POS tag c j appears after the POS tag c i in the entire corpus; f(c i ) is the frequency of the POS tag c i appearing in the entire corpus; and P T is the transition probability.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Definition 3. Given a tag set of size m(c 1 ,c 2 ,\u2026,c m ), the transition probability table T is an m by m matrix and given by:",
"sec_num": null
},
{
"text": "Definition 4. Given a tag set of size m(c 1 ,c 2 ,\u2026,c m ), the emission probability table E is an m by m matrix and given by:",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Definition 3. Given a tag set of size m(c 1 ,c 2 ,\u2026,c m ), the transition probability table T is an m by m matrix and given by:",
"sec_num": null
},
{
"text": "( , ) , ( ) ( , ) i j j f c c E i j i j f c E P c c = =",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Definition 3. Given a tag set of size m(c 1 ,c 2 ,\u2026,c m ), the transition probability table T is an m by m matrix and given by:",
"sec_num": null
},
{
"text": "where f(c i , c j ) is the frequency of the POS tag c i appearing before the POS tag c j in the entire corpus; f(c j ) is the frequency of POS tag c j appearing in the entire corpus; and P E is the emission probability.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Definition 3. Given a tag set of size m(c 1 ,c 2 ,\u2026,c m ), the transition probability table T is an m by m matrix and given by:",
"sec_num": null
},
{
"text": "Note that both T and E are constructed from the entire corpus and that we can look up these two tables easily when we consider the POS tags appears in POS tagging sequences. Now, when we look at a context window of size 7 (w 1 ,w 2 ,\u2026,w 7 ) and its POS tagging sequence (t 1 ,t 2 ,\u2026,t 7 ) , there are three types of probabilities we need to take into account.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Definition 3. Given a tag set of size m(c 1 ,c 2 ,\u2026,c m ), the transition probability table T is an m by m matrix and given by:",
"sec_num": null
},
{
"text": "The first one is the probability of the appearance of the POS tag t 4 of the multi-category word, which we can write as follows:",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Definition 3. Given a tag set of size m(c 1 ,c 2 ,\u2026,c m ), the transition probability table T is an m by m matrix and given by:",
"sec_num": null
},
{
"text": "4 4 4 4",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Definition 3. Given a tag set of size m(c 1 ,c 2 ,\u2026,c m ), the transition probability table T is an m by m matrix and given by:",
"sec_num": null
},
{
"text": "( ) ( is tagged as ) / ( )",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Definition 3. Given a tag set of size m(c 1 ,c 2 ,\u2026,c m ), the transition probability table T is an m by m matrix and given by:",
"sec_num": null
},
{
"text": "CX P t f t f \u03c9 \u03c9 =",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Definition 3. Given a tag set of size m(c 1 ,c 2 ,\u2026,c m ), the transition probability table T is an m by m matrix and given by:",
"sec_num": null
},
{
"text": "where f(w 4 ) is the frequency of the appearance of the multi-category word w 4 in the entire corpus and f(w 4 is tagged as t 4 ) is the frequency of the appearance where the word w 4 is tagged as t 4 in the entire corpus.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Definition 3. Given a tag set of size m(c 1 ,c 2 ,\u2026,c m ), the transition probability table T is an m by m matrix and given by:",
"sec_num": null
},
{
"text": "The second one is transition probability, which is the probability of the appearance of the POS tag t i+1 in the i+1 position after the POS tag t i in the i position and is shown as follows:",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Definition 3. Given a tag set of size m(c 1 ,c 2 ,\u2026,c m ), the transition probability table T is an m by m matrix and given by:",
"sec_num": null
},
{
"text": "( , 1) 1 1 ( , ) ( , )/( ) T T i i i i i i i P P t t f t t t + + + = =",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Definition 3. Given a tag set of size m(c 1 ,c 2 ,\u2026,c m ), the transition probability table T is an m by m matrix and given by:",
"sec_num": null
},
{
"text": "The last is emission probability, which is the probability of the appearance of the POS tag t i-1 in the i-1 position before the POS tag t i in the i position and is shown as follows:",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Definition 3. Given a tag set of size m(c 1 ,c 2 ,\u2026,c m ), the transition probability table T is an m by m matrix and given by:",
"sec_num": null
},
{
"text": "( 1, ) 1 1 ( , ) ( , )/( ) E T i i i i i i i P P t t f t t t \u2212 \u2212 \u2212 = =",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Definition 3. Given a tag set of size m(c 1 ,c 2 ,\u2026,c m ), the transition probability table T is an m by m matrix and given by:",
"sec_num": null
},
{
"text": "According to the above three probability formulas, we can build a seven-dimensional vector, where each dimension corresponds to one POS tag, respectively.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Definition 3. Given a tag set of size m(c 1 ,c 2 ,\u2026,c m ), the transition probability table T is an m by m matrix and given by:",
"sec_num": null
},
{
"text": "Definition 5: Given a multi-category word with a context window of size 7 and its POS tagging sequence, the dependency attribute vector V P of the multi-category word is defined as follows: ; ; E E E P P P P P P P P P = \u00d7 = \u00d7 = \u00d7",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Definition 3. Given a tag set of size m(c 1 ,c 2 ,\u2026,c m ), the transition probability table T is an m by m matrix and given by:",
"sec_num": null
},
{
"text": "Now we are ready to define the context vector of multi-category words.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Context Vector of Multi-Category Words",
"sec_num": null
},
{
"text": "Definition 6. Given a multi-category word with a context window of size l and its POS attribute matrix Y, position attribute vector V X , and dependency attribute vector V P , the context vector V S of the multi-category word is defined as follows:",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Context Vector of Multi-Category Words",
"sec_num": null
},
{
"text": "( ) S X P V V V Y \u03b1 \u03b2 = + \u00d7",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Context Vector of Multi-Category Words",
"sec_num": null
},
{
"text": "where \u03b1 and \u03b2 are the weights of the position attribute and the dependency attribute, respectively.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Context Vector of Multi-Category Words",
"sec_num": null
},
{
"text": "Note that we require \u03b1 + \u03b2 =1\uff0cand their optimal values are determined by experiments in our study. ",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Context Vector of Multi-Category Words",
"sec_num": null
},
{
"text": "Context vectors can be extended by using 4 to 7 preceding and following words as a substitute for the 3 preceding and following words we used in context windows and POS tagging sequences. We conducted experiments with a context window of size 3 to 7 on our sampled 1M-word training corpus performing closed tests. The experimental results are simultaneously evaluated in terms of both the precision of consistency checking and algorithm complexity. We plot the effect on precision and relative complexity of the number of preceding or following words in Figure 1 and Figure 2 , respectively. As shown in Figure 1 , the precision of consistency checking increases as we include more preceding and following words. In particular, the precision is improved by 1% when we use 7 preceding and following words.",
"cite_spans": [],
"ref_spans": [
{
"start": 554,
"end": 562,
"text": "Figure 1",
"ref_id": "FIGREF1"
},
{
"start": 567,
"end": 575,
"text": "Figure 2",
"ref_id": "FIGREF2"
},
{
"start": 604,
"end": 612,
"text": "Figure 1",
"ref_id": "FIGREF1"
}
],
"eq_spans": [],
"section": "Experiment on the Size of the Context Window",
"sec_num": "2.2"
},
{
"text": "When the context vector model is extended to contain 7 preceding and following words, the dimensionality of the position attribute vector, POS attribute vector, and dependency attribute vector increases from 7 to 15, which results in a significant change in run time. Suppose the run time of our algorithm with 3 preceding and following words is 1, then, extending the context vector to 15 dimensions, the change in run time of our algorithm is plotted in Figure 2 .",
"cite_spans": [],
"ref_spans": [
{
"start": 456,
"end": 464,
"text": "Figure 2",
"ref_id": "FIGREF2"
}
],
"eq_spans": [],
"section": "Experiment on the Size of the Context Window",
"sec_num": "2.2"
},
{
"text": "As shown in Figures 1 and 2 , the increase of complexity is much higher than that of precision when extending the number of preceding and following words. Hence, we chose 3 as the number of preceding and following words to form context windows and calculate context vectors.",
"cite_spans": [],
"ref_spans": [
{
"start": 12,
"end": 27,
"text": "Figures 1 and 2",
"ref_id": "FIGREF1"
}
],
"eq_spans": [],
"section": "Experiment on the Size of the Context Window",
"sec_num": "2.2"
},
{
"text": "When using our sampled 1M-word training corpus to conduct closed test, we found that consistency checking precision changes significantly with different values of \u03b1 and \u03b2 . Figure 3 shows the trend when \u03b1 varies from 0.1 to 0.9. As shown in Figure 3 , when \u03b1 =0.4 ( \u03b2 =0.6), consistency checking precision reaches the highest value. Hence, we used \u03b1 =0.4 and \u03b2 =0.6 in our experiments. At the same time, the results show \u03b1 and \u03b2 have an important effect on consistency checking precision, namely the position attribute vector and the dependency attribute vector are very important for describing the context vector of multi-category word, which is consistent with our experience.",
"cite_spans": [],
"ref_spans": [
{
"start": 173,
"end": 181,
"text": "Figure 3",
"ref_id": "FIGREF3"
},
{
"start": 241,
"end": 249,
"text": "Figure 3",
"ref_id": "FIGREF3"
}
],
"eq_spans": [],
"section": "Effect on Consistency Checking Precision of \u03b1 and \u03b2",
"sec_num": "2.3"
},
{
"text": "Our consistency checking algorithm is based on classification of context vectors of multi-category words. In particular, we first classify context vectors of each multi-category ",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Consistency Checking of POS Tagging",
"sec_num": "3."
},
{
"text": "After constructing context vectors for all multi-category words from their context windows and POS tagging sequences, the similarity of two context vectors is defined as the Euclidean Distance between the two vectors.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Similarity between Context Vectors of Multi-category Words",
"sec_num": "3.1"
},
{
"text": "1/ 2 2 1 ( , ) ( ) n i i i d x y x y x y = \u23a1 \u23a4 = \u2212 = \u2212 \u23a2 \u23a5 \u23a3 \u23a6 \u2211",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Similarity between Context Vectors of Multi-category Words",
"sec_num": "3.1"
},
{
"text": "where x and y are two arbitrary context vectors of n dimensions.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Similarity between Context Vectors of Multi-category Words",
"sec_num": "3.1"
},
{
"text": "Classification is a process to assign objects that need to be sorted into certain classes. In this paper, we used a popular classification method: the k-NN algorithm.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "k-NN Classification Algorithm",
"sec_num": "3.2"
},
{
"text": "Suppose we have c classes and a class (",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "k-NN Classification Algorithm",
"sec_num": "3.2"
},
{
"text": "( 1,2, , ) i i c \u03b8 = K ) has i N samples ( ( ) ( 1,2, ) i i j x j N = \u2026",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "k-NN Classification Algorithm",
"sec_num": "3.2"
},
{
"text": "). The idea of the k-NN algorithm is that for each unlabeled object x , compute the distances between x and",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "k-NN Classification Algorithm",
"sec_num": "3.2"
},
{
"text": "1 c i i M N = = \u2211 samples ( ( ) i j",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "k-NN Classification Algorithm",
"sec_num": "3.2"
},
{
"text": "x ) whose class is known, and select k samples (k nearest neighbors) with the smallest distance. This object x will be assigned to the class that contains the most samples in the k nearest neighbors.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "k-NN Classification Algorithm",
"sec_num": "3.2"
},
{
"text": "We now formally define the discriminant function and discriminant rule. Suppose 1 2 , , , c k k k \u2026 are the numbers of samples in the k nearest neighbors of the object x that belong to the classes 1 2 k , , , c \u03b8 \u03b8 \u03b8 \u03b8 , respectively. Define the discriminant function of the class i \u03b8 as",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "k-NN Classification Algorithm",
"sec_num": "3.2"
},
{
"text": "d i (x) = k i , i = 1,2,\u2026,c.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "k-NN Classification Algorithm",
"sec_num": "3.2"
},
{
"text": "Then, the discriminant rule determining the class of the object x can be defined as follows:",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "k-NN Classification Algorithm",
"sec_num": "3.2"
},
{
"text": "1,2,K, ( ) max ( ) m i m i c d x d x x \u03b8 = = \u21d2 \u2208",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "k-NN Classification Algorithm",
"sec_num": "3.2"
},
{
"text": "In this section, we describe the steps of our classification-based consistency checking algorithm in detail.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Consistency Checking Algorithm",
"sec_num": "3.3"
},
{
"text": "Step 1: Randomly sampling sentences containing multi-category words and checking their POS tagging manually. For each multi-category word, classifying the context vectors of the sampled POS tagging sequences, so that the context vectors that have the same POS for the multi-category word belonging to the same class. Let C ci (\"ci\" is a multi-category word)be the number of possible POS tags of the multi-category word ci.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Consistency Checking Algorithm",
"sec_num": "3.3"
},
{
"text": "Step 2: Give a context vector v of a multi-category word ci, calculating the distances between v and all the context vectors that contain the multi-category word ci in the training corpus, and selecting k context vectors whose distances are smaller than the others.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Consistency Checking Algorithm",
"sec_num": "3.3"
},
{
"text": "Step 3: According to the k-NN algorithm, checking the classes of the k nearest context vectors and classifying the vector v.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Consistency Checking Algorithm",
"sec_num": "3.3"
},
{
"text": "Step 4: Comparing the POS of the multi-category word ci in the class that the k-NN algorithm assigns v to and the POS tag of ci,. If they are the same, the POS tagging of the multi-category word ci is considered to be consistent, otherwise it is inconsistent.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Consistency Checking Algorithm",
"sec_num": "3.3"
},
{
"text": "The major disadvantage of this algorithm is the difficulty in selecting the value of k. If k is too small, the classification result is unstable; on the other hand, if k is too big, the classification deviation increases. Therefore, the selection of the value of k should depend on not only some general factors, but also the unique characteristics and structures of the data set. Hence, the best method of selecting k should be a strategy that has an ability to adapt different data sets. One such strategy is to try different values on the training set first, draw the performance curve, and select an optimal k value. Figure 4 shows the consistency checking precision values obtained with various k values in the k-NN algorithm. The precision values are closed test results on our 1M-word training corpus, and were obtained by using \u03b1 =0.4 and \u03b2 =0.6 in the context vector model. As shown in Figure 4 , when k continues to increase from 6, the precision remains the same. When k reaches 9, the precision starts declining. Our experiment with other \u03b1 and \u03b2 values also show similar trends. Hence, we chose k=6 in this paper. ",
"cite_spans": [],
"ref_spans": [
{
"start": 621,
"end": 629,
"text": "Figure 4",
"ref_id": "FIGREF6"
},
{
"start": 895,
"end": 903,
"text": "Figure 4",
"ref_id": "FIGREF6"
}
],
"eq_spans": [],
"section": "Consistency Checking Algorithm",
"sec_num": "3.3"
},
{
"text": "We evaluated our consistency checking algorithm on our 1.5M-word corpus (including the 1M-word training corpus) and conducted open and closed tests. The results are showed in Table 1 . The experimental results show two interesting trends. First, the precision and recall of our consistency checking algorithm are 87.20% and 92.67% in the closed test, respectively, and 85.24% and 85.84% in the open test, respectively. Compared to Zhang et al. 2004 , the precision of consistency checking is improved by 2~3%, and the recall is improved by 10%. The experimental results indicate that the context vector model has great improvements over the one used in Zhang et al. 2004 . Second, thanks to the considerable improvement of the recall, to some extent, our consistency checking algorithm prevents the occurrence of events with small probabilities in POS tagging.",
"cite_spans": [
{
"start": 431,
"end": 448,
"text": "Zhang et al. 2004",
"ref_id": "BIBREF9"
},
{
"start": 653,
"end": 670,
"text": "Zhang et al. 2004",
"ref_id": "BIBREF9"
}
],
"ref_spans": [
{
"start": 175,
"end": 182,
"text": "Table 1",
"ref_id": "TABREF2"
}
],
"eq_spans": [],
"section": "Experimental Results and Discussions",
"sec_num": "4."
},
{
"text": "After analyzing the experimental results we have come to the following conclusions:",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Conclusion and Future Research",
"sec_num": "5."
},
{
"text": "The proposed classification-based method is effective for consistency checking of POS tagging. It solves a difficult problem that easily appears in automatic POS tagging processes, and greatly improves the efficiency and quality of manual correction of automatic POS tagging.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Conclusion and Future Research",
"sec_num": "5."
},
{
"text": "The information in corpora can be considered as if it were in a multi-dimensional space with various attributes that have mutual influences. When conducting a consistency checking of POS tagging, it is incomplete if we only consider POS tagging and relationships between the POS without taking into account other attributes of words. We",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Conclusion and Future Research",
"sec_num": "5."
}
],
"back_matter": [
{
"text": "This research was partially supported by the National Natural Science Foundation of China No. 60473139, 60775041 and the Natural Science Foundation of Shanxi Province No. 20051034. ",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Acknowledgements",
"sec_num": null
},
{
"text": "may use other widely known attributes of words to improve the consistency checking of POS tagging.In the future, we plan to investigate more types of word attributes and incorporate linguistic and mathematical knowledge to develop better consistency checking models, which ultimately provide a better means of building high-quality Chinese corpora.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "annex",
"sec_num": null
}
],
"bib_entries": {
"BIBREF0": {
"ref_id": "b0",
"title": "The proofreading method study on consistence of segment and part-of-speech",
"authors": [
{
"first": "Y.-P",
"middle": [],
"last": "Du",
"suffix": ""
},
{
"first": "J.-H",
"middle": [],
"last": "Zheng",
"suffix": ""
}
],
"year": 2001,
"venue": "Computer Development and Application",
"volume": "14",
"issue": "10",
"pages": "16--18",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Du, Y.-P., and J.-H. Zheng, \"The proofreading method study on consistence of segment and part-of-speech,\" Computer Development and Application, 14(10), 2001, pp. 16-18.",
"links": null
},
"BIBREF1": {
"ref_id": "b1",
"title": "Categorical Am biguity and Information Content: A Corpus-based Study of Chinese",
"authors": [
{
"first": "C.-R",
"middle": [],
"last": "Huang",
"suffix": ""
},
{
"first": "R.-Y.",
"middle": [],
"last": "Chang",
"suffix": ""
}
],
"year": 2004,
"venue": "Journal of Chinese Language and Computing",
"volume": "14",
"issue": "2",
"pages": "157--165",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Huang, C.-R., and R.-Y. Chang, \"Categorical Am biguity and Information Content: A Corpus-based Study of Chinese,\" Journal of Chinese Language and Computing, 14(2), 2004, pp. 157-165.",
"links": null
},
"BIBREF2": {
"ref_id": "b2",
"title": "Chinese Part-of-Speech Tagging: One-at-a-Time or All-at-Once? Word-Based or Character-Based?",
"authors": [
{
"first": "H",
"middle": [
"T"
],
"last": "Ng",
"suffix": ""
},
{
"first": "K",
"middle": [
"L"
],
"last": "Jin",
"suffix": ""
}
],
"year": 2004,
"venue": "Proceedings of the 2004 Conference on Empirical Methods in Natural Language Processing",
"volume": "",
"issue": "",
"pages": "277--284",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Ng, H. T., and K. L. Jin, \"Chinese Part-of-Speech Tagging: One-at-a-Time or All-at-Once? Word-Based or Character-Based?,\" In Proceedings of the 2004 Conference on Empirical Methods in Natural Language Processing (EMNLP 2004), 2004, Barcelona, Spain, pp. 277-284.",
"links": null
},
"BIBREF3": {
"ref_id": "b3",
"title": "Analyzing Popular Clustering Algorithms from Different Viewpoints",
"authors": [
{
"first": "W.-N",
"middle": [],
"last": "Qian",
"suffix": ""
},
{
"first": "A.-Y",
"middle": [],
"last": "Zhou",
"suffix": ""
}
],
"year": 2002,
"venue": "Journal of Software",
"volume": "13",
"issue": "8",
"pages": "1382--1394",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Qian, W.-N., and A.-Y. Zhou, \"Analyzing Popular Clustering Algorithms from Different Viewpoints,\" Journal of Software, 13(8), 2002, pp. 1382-1394.",
"links": null
},
"BIBREF4": {
"ref_id": "b4",
"title": "Research on the method of automatic correction of chinese POS tagging",
"authors": [
{
"first": "Y.-L",
"middle": [],
"last": "Qian",
"suffix": ""
},
{
"first": "J.-H",
"middle": [],
"last": "Zheng",
"suffix": ""
}
],
"year": 2003,
"venue": "Journal Of Chinese Information Processing",
"volume": "18",
"issue": "2",
"pages": "30--35",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Qian, Y.-L., and J.-H. Zheng, \"Research on the method of automatic correction of chinese POS tagging,\" Journal Of Chinese Information Processing, 18(2), 2003, pp. 30-35.",
"links": null
},
"BIBREF5": {
"ref_id": "b5",
"title": "An approach to improving the quality of part-of-speech tagging of Chinese text",
"authors": [
{
"first": "Y.-L",
"middle": [],
"last": "Qian",
"suffix": ""
},
{
"first": "J.-H",
"middle": [],
"last": "Zheng",
"suffix": ""
}
],
"year": 2004,
"venue": "Proceedings of the 2004 IEEE International Conference on Information Technology:Coding and Computing",
"volume": "",
"issue": "",
"pages": "183--187",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Qian, Y.-L., and J.-H. Zheng, \"An approach to improving the quality of part-of-speech tagging of Chinese text,\" In Proceedings of the 2004 IEEE International Conference on Information Technology:Coding and Computing (ITCC 2004), 2004, USA, pp. 183-187.",
"links": null
},
"BIBREF6": {
"ref_id": "b6",
"title": "Research and practice on the machine proofreading of the tagged corpora",
"authors": [
{
"first": "W.-G",
"middle": [],
"last": "Qu",
"suffix": ""
},
{
"first": "X.-H",
"middle": [],
"last": "Chen",
"suffix": ""
}
],
"year": 2003,
"venue": "Proceeding of the 7 th National Computation Linguistics (JSCL'03)",
"volume": "",
"issue": "",
"pages": "318--324",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Qu, W.-G., and X.-H. Chen, \"Research and practice on the machine proofreading of the tagged corpora,\" In Proceeding of the 7 th National Computation Linguistics (JSCL'03), 2003, Beijing, China, pp. 318-324.",
"links": null
},
"BIBREF7": {
"ref_id": "b7",
"title": "Analysing on the words classified hard in pos tagging",
"authors": [
{
"first": "H.-B",
"middle": [],
"last": "Xing",
"suffix": ""
}
],
"year": 1999,
"venue": "Proceedings of the 5th National Computational Linguistics (JSCL'99)",
"volume": "",
"issue": "",
"pages": "187--192",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Xing, H.-B., \"Analysing on the words classified hard in pos tagging,\" In Proceedings of the 5th National Computational Linguistics (JSCL'99), 1999, Beijing, China, pp. 187-192.",
"links": null
},
"BIBREF8": {
"ref_id": "b8",
"title": "Building a Large-Scale Annotated Chinese Corpus",
"authors": [
{
"first": "N",
"middle": [],
"last": "Xue",
"suffix": ""
},
{
"first": "F.-D",
"middle": [],
"last": "Chiou",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Palmer",
"suffix": ""
}
],
"year": 2002,
"venue": "Proceedings of the 19th international conference on Computational linguistics",
"volume": "",
"issue": "",
"pages": "1--8",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Xue, N., F.-D. Chiou, and M. Palmer, \"Building a Large-Scale Annotated Chinese Corpus,\" In Proceedings of the 19th international conference on Computational linguistics, 2002, Taipei, Taiwan, pp. 1-8.",
"links": null
},
"BIBREF9": {
"ref_id": "b9",
"title": "The Inspecting Method Study on Consistency of POS Tagging of Corpus",
"authors": [
{
"first": "H",
"middle": [],
"last": "Zhang",
"suffix": ""
},
{
"first": "J.-H",
"middle": [],
"last": "Zheng",
"suffix": ""
}
],
"year": 2004,
"venue": "Journal Of Chinese Information Processing",
"volume": "18",
"issue": "5",
"pages": "11--16",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Zhang, H., and J.-H. Zheng, \"The Inspecting Method Study on Consistency of POS Tagging of Corpus,\" Journal Of Chinese Information Processing, 18(5), 2004, pp. 11-16.",
"links": null
},
"BIBREF10": {
"ref_id": "b10",
"title": "A Study on Consistency Checking Method of 169",
"authors": [],
"year": null,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "A Study on Consistency Checking Method of 169",
"links": null
},
"BIBREF11": {
"ref_id": "b11",
"title": "Speech Tagging for Chinese Corpora",
"authors": [],
"year": null,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Part-Of-Speech Tagging for Chinese Corpora",
"links": null
},
"BIBREF12": {
"ref_id": "b12",
"title": "POS Tagging Chinese Corpus Based on Statistics and Rules",
"authors": [
{
"first": "M",
"middle": [],
"last": "Zhang",
"suffix": ""
},
{
"first": "S",
"middle": [],
"last": "Li",
"suffix": ""
}
],
"year": 1998,
"venue": "Journal of Software",
"volume": "9",
"issue": "2",
"pages": "134--138",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Zhang, M., and S. Li, \"POS Tagging Chinese Corpus Based on Statistics and Rules,\" Journal of Software, 9(2), 1998, pp. 134-138.",
"links": null
}
},
"ref_entries": {
"FIGREF1": {
"text": "Effect on precision of the number of preceding and following words.",
"num": null,
"uris": null,
"type_str": "figure"
},
"FIGREF2": {
"text": "Effect on complexity of the number of preceding and following words.",
"num": null,
"uris": null,
"type_str": "figure"
},
"FIGREF3": {
"text": "Effect on consistency checking precision of \u03b1 and \u03b2 .",
"num": null,
"uris": null,
"type_str": "figure"
},
"FIGREF4": {
"text": "Study on Consistency Checking Method of 165 Part-Of-Speech Tagging for Chinese Corpora word in the training corpus, and then we conduct the consistency checking of POS tagging based on classification results.",
"num": null,
"uris": null,
"type_str": "figure"
},
"FIGREF6": {
"text": "Effect on precision of k in the k-NN algorithm.",
"num": null,
"uris": null,
"type_str": "figure"
},
"FIGREF7": {
"text": "Study on Consistency Checking Method of 167 Part-Of-Speech Tagging for Chinese Corpora",
"num": null,
"uris": null,
"type_str": "figure"
},
"TABREF0": {
"html": null,
"type_str": "table",
"text": "Speech Tagging for Chinese CorporaIn the same way, if the size of a context window is 15, the position attribute vector can be described as follows:",
"content": "<table><tr><td>A Study on Consistency Checking Method of</td><td>161</td></tr><tr><td>Part-Of-((1/190),(2/190),(4/190),(8/190),(16/190),(32/190),</td><td/></tr><tr><td>(64/190),(32/190),(16/190),(8/190),(4/190),(2/190),(1/190))</td><td/></tr></table>",
"num": null
},
"TABREF2": {
"html": null,
"type_str": "table",
"text": "",
"content": "<table><tr><td>Test corpora</td><td/><td colspan=\"2\">Test type</td><td/><td>Number of multi-category words</td><td>Number of the true inconsistencies</td><td>Number of the identified (True) inconsistencies</td><td>Recall (%)</td><td>Precision (%)</td></tr><tr><td>1M word</td><td colspan=\"4\">closed</td><td>127,210</td><td>1,147</td><td>1,219(1063)</td><td>92.67</td><td>87.20</td></tr><tr><td>500K word</td><td/><td colspan=\"2\">open</td><td/><td>64,467</td><td>579</td><td>583(497)</td><td>85.84</td><td>85.24</td></tr><tr><td colspan=\"2\">Recall</td><td>=</td><td colspan=\"4\">Number of the identified true inconsistencies Number of the true inconsistencies</td></tr><tr><td colspan=\"3\">Precision</td><td>=</td><td colspan=\"3\">Number of the identified true inconsistencies Number of the identified inconsistencies</td></tr></table>",
"num": null
}
}
}
} |