File size: 54,609 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 |
{
"paper_id": "O07-1003",
"header": {
"generated_with": "S2ORC 1.0.0",
"date_generated": "2023-01-19T08:08:27.904528Z"
},
"title": "Korean-Chinese Cross-Language Information Retrieval Based on Extension of Dictionaries and Transliteration",
"authors": [
{
"first": "Yu-Chun",
"middle": [],
"last": "Wang",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "Academia Sinica",
"location": {
"country": "Taiwan"
}
},
"email": ""
},
{
"first": "Richard",
"middle": [],
"last": "Tzong",
"suffix": "",
"affiliation": {},
"email": ""
},
{
"first": "Han",
"middle": [],
"last": "Tsai",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "Academia Sinica",
"location": {
"country": "Taiwan"
}
},
"email": "thtsai@iis.sinica.edu.tw"
},
{
"first": "Hsu-Chun",
"middle": [],
"last": "Yen",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "National Taiwan University",
"location": {
"country": "Taiwan"
}
},
"email": "yen@cc.ee.ntu.edu.tw"
},
{
"first": "Wen-Lian",
"middle": [],
"last": "Hsu",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "Academia Sinica",
"location": {
"country": "Taiwan"
}
},
"email": "hsu@iis.sinica.edu.tw"
}
],
"year": "",
"venue": null,
"identifiers": {},
"abstract": "This paper describes our Korean-Chinese cross-language information retrieval system. Our system uses a bilingual dictionary to perform query translation. We expand our bilingual dictionary by extracting words and their translations from the Wikipedia site, an online encyclopedia. To resolve the problem of translating Western people's names into Chinese, we propose a transliteration mapping method. We translate queries form Korean query to Chinese by using a co-occurrence method. When evaluating on the NTCIR-6 test set, the performance of our system achieves a mean average precision (MAP) of 0.1392 (relax score) for title query type and 0.1274 (relax score) for description query type. \u6458\u8981 \u672c\u6587\u63cf\u8ff0\u6211\u5011\u6240\u63d0\u51fa\u4e4b\u97d3\u4e2d\u96d9\u8a9e\u8de8\u8a9e\u8a00\u6aa2\u7d22\u7cfb\u7d71\u3002\u6211\u5011\u63a1\u7528\u97d3\u4e2d\u96d9\u8a9e\u8fad\u5178\u9032\u884c\u554f\u984c\u4e4b\u7ffb\u8b6f\uff0c\u4e26 \u5229\u7528\u7dda\u4e0a\u7dad\u57fa\u767e\u79d1\u4ee5\u53ca\u97d3\u570b Naver \u7db2\u7ad9\u4f86\u64f4\u589e\u6211\u5011\u96d9\u8a9e\u8fad\u5178\u7684\u8986\u84cb\u7387\u3002\u6b64\u5916\uff0c\u91dd\u5c0d\u97d3\u6587\u4e2d\u897f \u65b9\u4eba\u540d\u7684\u7ffb\u8b6f\uff0c\u6211\u5011\u63d0\u51fa\u4e00\u97f3\u8b6f\u5c0d\u61c9\u7684\u641c\u5c0b\u65b9\u6cd5\u3002\u5c0d\u65bc\u97d3\u4e2d\u7ffb\u8b6f\u6642\u7684\u6b67\u7fa9\u6027\u554f\u984c\uff0c\u6211\u5011\u63a1\u7528 Mutual Information \u65b9\u6cd5\u4f86\u89e3\u6c7a\u3002\u6211\u5011\u4f7f\u7528 NTCIR-6 \u4e4b test set \u6e2c\u8a66\u6211\u5011\u97d3\u4e2d\u8de8\u8a9e\u8a00\u6aa2\u7d22\u7cfb\u7d71 \u4e4b\u6548\u7387\uff0c\u5176\u4f7f\u7528\u6a19\u984c\u90e8\u5206\u9032\u884c\u67e5\u8a62\u6642\u4e4b Mean average precision (MAP) \u4e4b\u7d50\u679c\u70ba 0.1392\uff1b\u4f7f \u7528\u6558\u8ff0\u90e8\u5206\u9032\u884c\u67e5\u8a62\u6642\u4e4b MAP \u70ba 0.1274\u3002",
"pdf_parse": {
"paper_id": "O07-1003",
"_pdf_hash": "",
"abstract": [
{
"text": "This paper describes our Korean-Chinese cross-language information retrieval system. Our system uses a bilingual dictionary to perform query translation. We expand our bilingual dictionary by extracting words and their translations from the Wikipedia site, an online encyclopedia. To resolve the problem of translating Western people's names into Chinese, we propose a transliteration mapping method. We translate queries form Korean query to Chinese by using a co-occurrence method. When evaluating on the NTCIR-6 test set, the performance of our system achieves a mean average precision (MAP) of 0.1392 (relax score) for title query type and 0.1274 (relax score) for description query type. \u6458\u8981 \u672c\u6587\u63cf\u8ff0\u6211\u5011\u6240\u63d0\u51fa\u4e4b\u97d3\u4e2d\u96d9\u8a9e\u8de8\u8a9e\u8a00\u6aa2\u7d22\u7cfb\u7d71\u3002\u6211\u5011\u63a1\u7528\u97d3\u4e2d\u96d9\u8a9e\u8fad\u5178\u9032\u884c\u554f\u984c\u4e4b\u7ffb\u8b6f\uff0c\u4e26 \u5229\u7528\u7dda\u4e0a\u7dad\u57fa\u767e\u79d1\u4ee5\u53ca\u97d3\u570b Naver \u7db2\u7ad9\u4f86\u64f4\u589e\u6211\u5011\u96d9\u8a9e\u8fad\u5178\u7684\u8986\u84cb\u7387\u3002\u6b64\u5916\uff0c\u91dd\u5c0d\u97d3\u6587\u4e2d\u897f \u65b9\u4eba\u540d\u7684\u7ffb\u8b6f\uff0c\u6211\u5011\u63d0\u51fa\u4e00\u97f3\u8b6f\u5c0d\u61c9\u7684\u641c\u5c0b\u65b9\u6cd5\u3002\u5c0d\u65bc\u97d3\u4e2d\u7ffb\u8b6f\u6642\u7684\u6b67\u7fa9\u6027\u554f\u984c\uff0c\u6211\u5011\u63a1\u7528 Mutual Information \u65b9\u6cd5\u4f86\u89e3\u6c7a\u3002\u6211\u5011\u4f7f\u7528 NTCIR-6 \u4e4b test set \u6e2c\u8a66\u6211\u5011\u97d3\u4e2d\u8de8\u8a9e\u8a00\u6aa2\u7d22\u7cfb\u7d71 \u4e4b\u6548\u7387\uff0c\u5176\u4f7f\u7528\u6a19\u984c\u90e8\u5206\u9032\u884c\u67e5\u8a62\u6642\u4e4b Mean average precision (MAP) \u4e4b\u7d50\u679c\u70ba 0.1392\uff1b\u4f7f \u7528\u6558\u8ff0\u90e8\u5206\u9032\u884c\u67e5\u8a62\u6642\u4e4b MAP \u70ba 0.1274\u3002",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Abstract",
"sec_num": null
}
],
"body_text": [
{
"text": "The contents of whole Internet are growing explosively due to the improvement of the computer and web technology. Besides English, the web pages written in other languages also increase tremendously. In order to get the useful information from the Internet, many advanced modern search engines are developed, like Google 1 , Yahoo 2 , AltaVista 3 , and so on. However, for the users that do not have any knowledge about other languages, it is impossible to get the information in other languages by current single-language web search engines.",
"cite_spans": [
{
"start": 345,
"end": 346,
"text": "3",
"ref_id": "BIBREF2"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Instroduction",
"sec_num": "1"
},
{
"text": "Therefore, the research of cross language information retrieval (CLIR) is rising quickly. Cross language information retrieval systems allow the users to input the key words in their own languages and then the systems will retrieve the relevant documents written in the other language that the users want to search based on the queries the users inputted.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Instroduction",
"sec_num": "1"
},
{
"text": "There are many different approaches of CLIR. The first is the translation methods. There are two kinds of approaches usually adopted: translation approach and statistical approach. The translation approach uses the bilingual dictionaries, ontology, or thesaurus to translate either queries or documents. The statistical approach uses pre-constructed bilingual corpora to extract the cross-lingual associations without any language translation methods [1] [2] [3] . The translation approach is restricted with the coverage and the precision of the dictionaries. The statistical approach can extract bilingual lexicons automatically; however, it bases on a well-constructed and large-scaled bilingual corpus which requires a lot of human effort.",
"cite_spans": [
{
"start": 451,
"end": 454,
"text": "[1]",
"ref_id": "BIBREF0"
},
{
"start": 455,
"end": 458,
"text": "[2]",
"ref_id": "BIBREF1"
},
{
"start": 459,
"end": 462,
"text": "[3]",
"ref_id": "BIBREF2"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Instroduction",
"sec_num": "1"
},
{
"text": "In translation approach, there are two different targets to do the translation. One is document translation; the other is query translation. The document translation approach is to translate all the documents in the collection from the target language to the source language the users use. Then, while the users give the query, the system will do a monolingual information retrieval. The query translation translates the query in the source language that user inputted into the target language and then retrieval the documents which is written in the target language. The document translation approach is possible if there exists a high quality machine translation system. [4, 5] However, the document translation approach is not very practical when the documents are not stable or can be updated frequently, like the web text retrieval.",
"cite_spans": [
{
"start": 673,
"end": 676,
"text": "[4,",
"ref_id": "BIBREF3"
},
{
"start": 677,
"end": 679,
"text": "5]",
"ref_id": "BIBREF4"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Instroduction",
"sec_num": "1"
},
{
"text": "In this paper, we propose a Korean-Chinese cross-language information retrieval system. We adopt the query-translation approach because it is effective. Moreover, the translation method, which is dictionary-based, does not involve a great deal of work. In CLIR, the most serious problem is that unknown words cannot be translated correctly. To resolve the problem, we utilize Wikipedia, an online encyclopedia, to expand our dictionary to make higher coverage of vocabulary. Another difficult issue involves translating Western people's names written in Korean into Chinese. As a solution, we propose a transliteration mapping method to deal with the problem.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Instroduction",
"sec_num": "1"
},
{
"text": "The remainder of the paper is organized as follows. In Section 2, we give an overview of our system and describe its implementation, including the translation and indexing methods adopted. In Section 3, we detail the evaluation results of our CLIR system based on the topics and the document collections provided by NTCIR CLIR task, and discuss the effectiveness of our method, as well as some problems that have to be solved. Finally, in Section 4, we present our conclusions and indicate the direction of our future work. Figure 1 shows the architecture of our CLIR system. It is comprised of four stages. First, a Korean query is chunked into several key terms, which are then translated into Chinese by three dictionaries. In the third stage, we disambiguate the translated terms and transform them into a Lucene query. Finally, the query is sent to the Lucene IR engine and the answer is retrieved.",
"cite_spans": [],
"ref_spans": [
{
"start": 524,
"end": 532,
"text": "Figure 1",
"ref_id": null
}
],
"eq_spans": [],
"section": "Instroduction",
"sec_num": "1"
},
{
"text": "Unlike English, Korean written texts do not have word delimiters. Spaces in Korean sentences separate eojeols, which are composed of a noun and a postposition, or a verb stem and a verb ending. Therefore, Korean text has to be segmented. There are two types of queries that the users might make. One is composed of several key words; the other is an natural language sentence. Therefore, we use two different segmentation methods to deal with these two query types separately.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Query Processing",
"sec_num": "2.1"
},
{
"text": "Due to the characteristics of the Korean language, the keyword-typed queries written in Korean are comprised mainly of nouns. We use spaces to split the title into several eojeols, and Figure 1 : System Architecture of Our CLIR System then remove the postpositions at the end of the eojeols according to our predefined rules. For the natural language sentence of a Korean query, we use the KLT Term Extractor 4 , developed by Kookmin University in Korea. KLT term extractor will do the word segmentation to extract vital key words which are useful for information retrieval and remove stop words.",
"cite_spans": [],
"ref_spans": [
{
"start": 185,
"end": 193,
"text": "Figure 1",
"ref_id": null
}
],
"eq_spans": [],
"section": "Query Processing",
"sec_num": "2.1"
},
{
"text": "Due to copyright restrictions, we use the free online Korean-Chinese dictionary provided by the Daum Korean web site 5 . We send the key terms obtained in the query processing stage to the online dictionary. However, as a general bilingual dictionary is not suitable for proper nouns, we use Wikipedia 6 , an online encyclopedia, to expand our dictionary. In Wikipedia, an item might contain inter-language links to the same item in Wikipedia written in other languages. Therefore, we send a Korean term to Korean Wikipedia. If it contains an inter-language link to Chinese Wikipedia, we can find the corresponding Chinese word. This method is very efficient because it yields accurate Chinese translations of Korean words.",
"cite_spans": [
{
"start": 117,
"end": 118,
"text": "5",
"ref_id": "BIBREF4"
},
{
"start": 302,
"end": 303,
"text": "6",
"ref_id": "BIBREF5"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Bilingual Dictionary Translation",
"sec_num": "2.2.1"
},
{
"text": "The Daum Korean-Chinese dictionary is written in simplified Chinese, as are many pages in Chinese Wikipedia. We use a simple mapping table to convert simplified Chinese characters to traditional Chinese characters.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Bilingual Dictionary Translation",
"sec_num": "2.2.1"
},
{
"text": "If some terms cannot be found in the Daum dictionary or Wikipedia, we apply the maximal matching algorithm to split a long term into several shorter terms. Then, the shorter terms are sent to the Daum dictionary and Wikipedia to search for Chinese translations.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Bilingual Dictionary Translation",
"sec_num": "2.2.1"
},
{
"text": "The person names may often appear in the query. Although Wikipedia contains many famous people's names around the world, some people's names are still excluded. Therefore, it is necessary to deal with this person name translation problem. Unlike Korean-English or Korean-Japanese CLIR, transliteration methods are not appropriate for Korean-Chinese CLIR because so many Chinese characters have the same pronunciation in Korean. Besides, to translate Japanese personal names, Korean uses the Hangul alphabet to pronounce the names of Japanese people; however, Chinese uses original Chinese characters with Mandarin pronunciation, instead of Japanese pronunciation of Chinese characters. Thus, transliteration methods are not useful in this context. To solve the problem, we use Naver People Search 7 , a database containing the basic profiles of famous people, including their original names. We can submit person names in Korean to Naver people search and get their original names. If the original name is composed of Chinese characters, it is clearly Chinese, Japanese, or Korean; therefore, we can send it to next stage directly, i.e., the disambiguation stage. If, however, the original name is in English, we use the English name translation table provided by Taiwan's Central News Agency (CNA) 8 to translate it into Chinese and the proceed to the next stage.",
"cite_spans": [
{
"start": 1299,
"end": 1300,
"text": "8",
"ref_id": "BIBREF7"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Person Name Translation",
"sec_num": "2.2.2"
},
{
"text": "In the past, the Korean language adopted many Chinese words. More than half of its vocabulary comprises Chinese words. Now, however, Koreans use Hangul, an alphabet writing system, instead of Chinese characters, which is an ideograph writing system. As a result, many different Chinese loanwords have the same pronunciation when written in the Hangul alphabet. For example, the four different Chinese loanwords with different meanings: \"\u7406\u60f3\" (ideal), \"\u4ee5\u4e0a\" (above), \"\u7570\u5e38\" (unusual), and \"\u7570\u72c0\" (indisposition) are written in the same way as the Hangul word \"\uc774\uc0c1\" because their pronunciation is the same in Korean. This creates a very serious ambiguity problem when Korean is translated into Chinese. Therefore, choosing the correct translation term among translation candidates is important.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Term Disambiguation",
"sec_num": "2.3"
},
{
"text": "For each term in a given query Q, there may be several possible translation candidates. To select the best translation term among all the candidates, we must not only consider the original query term qt but also consider all the other terms in Q and their translation candidates. We denote the j-th translation candidate for the i-th term qt i in Q as tc ij . We adopt the mutual information score (MI score) [6] to evaluate the co-relation between the tc ij and all translation candidates of all the other terms in Q. The MI score of tc ij given Q is calculated as follows:",
"cite_spans": [
{
"start": 409,
"end": 412,
"text": "[6]",
"ref_id": "BIBREF5"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Term Disambiguation",
"sec_num": "2.3"
},
{
"text": "MI score(tc ij |Q) = |Q| \u2211 x=1,x\u0338 =i Z(qtx) \u2211 y=1 P r(tc ij , tc xy ) P r(tc ij )P r(tc xy )",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Term Disambiguation",
"sec_num": "2.3"
},
{
"text": ",",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Term Disambiguation",
"sec_num": "2.3"
},
{
"text": "where Z(qt x ) is the number of translation candidates of the x-th query term qt x ; P r(tc ij , tc xy ) is the probability that tc ij and tc xy co-occur in the same sentence; and P r(tc ij ) is the probability of tc ij . The values of the probabilities are obtained from Chinese Information Retrieval Benchmark (CIRB) Chinese corpus which is provided by NTCIR CLIR task [7] . The higher the translation candidate's MI score is, the higher weight is assigned to it in the retrieval module.",
"cite_spans": [
{
"start": 371,
"end": 374,
"text": "[7]",
"ref_id": "BIBREF6"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Term Disambiguation",
"sec_num": "2.3"
},
{
"text": "The Chinese documents we use is CIRB 4.0 documents which is provided by NTCIR. The CIRB 4.0 documents are pre-processed to remove noise and then segmented by CKIP AutoTag [8] to obtain words and part-of-speech (POS). We use Lucene 9 , an open source information retrieval engine, to index Chinese documents. Our index is based on Chinese characters.",
"cite_spans": [
{
"start": 171,
"end": 174,
"text": "[8]",
"ref_id": "BIBREF7"
},
{
"start": 231,
"end": 232,
"text": "9",
"ref_id": "BIBREF8"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Chinese Document Indexing",
"sec_num": "2.4"
},
{
"text": "After processing a Korean query into several terms and translating it into Chinese, we transform the Chinese terms into a Lucene Query. Different Chinese terms are separated by a space, which means an \"OR\" operator in the Lucene format. If a term has different translation candidates, the weight of the candidate with highest mutual information score will be increased by 1 by the boost operator. The other candidates are boosted by a weight that is the reciprocal of the total number of candidates. The boost operation affects the ranking of the documents the Lucene returns. The default boost value of each terms is 1, and we decrease the weight of the candidates with lower mutual information score to make them not affect the ranking so much.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Lucene Queries",
"sec_num": "2.5"
},
{
"text": "In order to evaluate our CLIR system, we use the topics and the document collections which is provided by NTCIR-6 CLIR task [7] . The topics contains 50 Korean queries composed of four parts: title, description, narration, and keywords. We use these topics as the queries that users inputted in our system. The main metric to evaluate the performance of information retrieval is Mean Average Precision (MAP) [9] . Average precision is based on the whole list of documents returned by the system and emphasizes returning more relevant documents earlier. The Mean Average Precision is the mean value of the average precisions computed for each query. Besides, R-precision [10] is also a good metric which is the precision among the front of R relevant documents.",
"cite_spans": [
{
"start": 124,
"end": 127,
"text": "[7]",
"ref_id": "BIBREF6"
},
{
"start": 408,
"end": 411,
"text": "[9]",
"ref_id": "BIBREF8"
},
{
"start": 670,
"end": 674,
"text": "[10]",
"ref_id": null
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Evaluation and Analysis",
"sec_num": "3"
},
{
"text": "There are two kinds of relevance judgments: Rigid and Relax. A document is rigid relevant if it is highly relevant; a document is relax relevant if it is highly relevant or partial relevant. Our evaluation is based on the 50 topics which is selected by NTCIR-6 CLIR task to compute among all 140 topics they provided.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Evaluation and Analysis",
"sec_num": "3"
},
{
"text": "In order to evaluate the effectiveness of our CLIR system, we build a monolingual Chinese IR system for comparison. NTCIR-6 CLIR test set also contains the Chinese topics which meanings are the same as Korean ones. We use these Chinese topics as queries and apply CKIP AutoTag to do Chinese word segmentation and remove Chinese stop words. Then, we use Lucene search engine we use in our CLIR system to retrieve related Chinese documents.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Evaluation and Analysis",
"sec_num": "3"
},
{
"text": "We do the four different runs:",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Evaluation and Analysis",
"sec_num": "3"
},
{
"text": "\u2022 KC-title-run: a run using a Korean title field to retrieve Chinese documents.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Evaluation and Analysis",
"sec_num": "3"
},
{
"text": "\u2022 KC-description-run: a run using a Korean description field to retrieve Chinese documents.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Evaluation and Analysis",
"sec_num": "3"
},
{
"text": "\u2022 CC-title-run: a monolingual Chinese run using Chinese title field to retrieve Chinese documents. \u2022 CC-description-run: a monolingual Chinese run using Chinese description field to retrieve Chinese documents. Table 1 shows the performance of our Korean-Chinese CLIR system and the monolingual Chinese IR system. The performance of Korean-Chinese CLIR is not as good as that of Chinese monolingual IR. We have investigated why it is difficult to retrieve high precision answers to some queries.",
"cite_spans": [],
"ref_spans": [
{
"start": 210,
"end": 217,
"text": "Table 1",
"ref_id": "TABREF1"
}
],
"eq_spans": [],
"section": "Evaluation and Analysis",
"sec_num": "3"
},
{
"text": "We use a general bilingual dictionary and Wikipedia to translate most of the words in 50 topics provided by NTCIR-6 CLIR task. Although we have used Wikipedia to expand our dictionary, there are some problems that cause translations to fail. The first problem is that there are still some unknown words. For example, the word \"\ubc30\uc544\" (embryo) is not listed in the dictionaries. The other problem is that the dictionaries do not always have the proper translation candidates of the words and terms in queries. For instance, the word \"\uac10\uccad\" (monitor) is not translated correctly because the dictionary lacks the correct translation and provides another translation instead, i.e., \"\u7d3a\u9752\" (deep blue). Also, the word \"\uc554\" (cancer) in one topic is translated as \"\u5ca9\" (rock), \"\u5eb5\" (nunnery), and \"\u96cc\" (female), but no correct translation, i.e., \"\u764c\" (cancer).",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Problems of Bilingual Dictionaries",
"sec_num": "3.1"
},
{
"text": "The Daum Korean-Chinese dictionary that we use was written people studying Mainland Chinese, i.e., Pinyin. However, the CIRB 4.0 document collection contains Taiwanese newspapers. Taiwanese people use traditional Chinese characters, whereas Mainland Chinese people use simplified characters. Besides the difference in characters, the vocabulary and grammar used in Taiwan and China are slightly different. The differences between Taiwanese Chinese and Mainland Chinese can make IR difficult.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Different Phraseology Used in Taiwan and China",
"sec_num": "3.2"
},
{
"text": "The following are some examples of the difficulties we face. The term \"\ud734\ub300\ud3f0\" (mobile phone) is translated into Mainland Chinese word as \"\u79fb\u52d5\u96fb\u8a71\" (the phone that can move); however, the correct word used in Taiwan is \"\u624b\u6a5f\" (the machine held in the hand). The word \"\uc720\uc804\uc790\" (gene) is translated to \"\u907a\u50b3\u5b50\" (the factor of heredity) , not to correct word \"\u57fa\u56e0\" (the Mandarin transliteration of the English word \"gene\") used in Taiwan. The word \"\uc778\ud130\ub137\" (internet) in some topics is translated to \"\u4e92\u806f\u7db2\" (the net connecting to each other) , but the correct word used in Taiwan is \"\u7db2\u969b\u7db2\u8def\" (cyber network).",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Different Phraseology Used in Taiwan and China",
"sec_num": "3.2"
},
{
"text": "If a term is not defined in our dictionaries, we split it into several shorter terms by the maximal matching algorithm discussed in Section 2.2.1. In some cases, however, the algorithm do not segment a term correctly. For example, for the term \"\ube44\uc811\ucd09\ud615\"(contactless), the correct segmentation is \ube44(not)-\uc811\ucd09(contact)-\ud615(type). However, it is segmented as \ube44\uc811-\ucd09-\ud615 so that the wrong word, \"\ube44\uc811\"(convalescing), is retrieved.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "The Limitations of Maximal Matching Algorithm",
"sec_num": "3.3"
},
{
"text": "In some topics, different expressions used in Korean and Chinese may cause translation problems. In one of the topic, the word \"10\ub300\" refers to people aged between 10 and 19. Similarly, \"20\ub300\" means people aged from 20 to 29. Therefore, the corresponding translation of the word \"10\ub300\" in this topic is \"\u9752\u5c11\u5e74\" (teenager). However, our system translates the numbers and the Hangul characters separately so that the final translation is \"10\u4ee3\" (ten generations). This is a semantic problem that our system has difficulty coping with.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Different Expressions Used in Korean and Chinese",
"sec_num": "3.4"
},
{
"text": "Another problem relates to abbreviations used in Chinese. For instance, in another topic, \"\uc65c\uad6d\uc778 \ub178\ub3d9\uc790\" (foreign worker) is translated into \"\u5916\u570b\u4eba\u52de\u5de5\" (foreign worker) by our system. However, in Taiwanese newspapers, the abbreviation \"\u5916\u52de\" , which is composed of the first characters of the two words : \"\u5916\u570b\u4eba\" (foreigner) and \"\u52de\u5de5\" (worker), is used more frequently. Our translation in one of the topic for the phrases \"\uc6d0\uc790\ub2a5 \ubc18\ub300\" is \"\u53cd\u5c0d \u6838\u80fd\" (anti-nuclear), but the abbreviation \"\u53cd\u6838\" is frequently used.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Different Expressions Used in Korean and Chinese",
"sec_num": "3.4"
},
{
"text": "We have described our Korean-Chinese CLIR system. It is based on a query-translation approach and uses a general Korean-Chinese dictionary and Wikipedia to translate words and terms. To obtain person names, we use the Naver people search website and the CNA transliteration table to translate the names.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Conclusions and Future Works",
"sec_num": "4"
},
{
"text": "We have evaluated the performance of our Korean-Chinese CLIR system with the Korean topics and the Chinese document collection which is provided by NTCIR-6 CLIR task. Our translation method is effective, but there are still some cases where the precision is low. We believe the problems are due to the limitations of the dictionaries, the different phraseology used in Taiwan and China, and the expressions used in Chinese and Korean.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Conclusions and Future Works",
"sec_num": "4"
},
{
"text": "In our future work, we will apply a Chinese thesaurus to overcome the problem of different Chinese phraseology and use more bilingual dictionaries to reduce the number of unknown words. We will also incorporate a query expansion method into our CLIR system to improve its precision.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Conclusions and Future Works",
"sec_num": "4"
},
{
"text": "http://www.google.com 2 http://www.yahoo.com 3 http://www.altavista.com",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
},
{
"text": "http://nlp.kookmin.ac.kr/HAM/kor/index.html 5 http://cndic.daum.net 6 http://www.wikipedia.org",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
},
{
"text": "http://people.naver.com 8 http://client.cna.com.tw/name/",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
},
{
"text": "http://lucene.apache.org/",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
}
],
"back_matter": [],
"bib_entries": {
"BIBREF0": {
"ref_id": "b0",
"title": "Automatic cross-language retrieval using latent semantic indexing",
"authors": [
{
"first": "S",
"middle": [],
"last": "Dumais",
"suffix": ""
},
{
"first": "T",
"middle": [],
"last": "Letsche",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Littman",
"suffix": ""
},
{
"first": "T",
"middle": [],
"last": "Landauer",
"suffix": ""
}
],
"year": 1997,
"venue": "AAAI Symposium on CrossLanguage Text and Speech Retrieval",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "S. Dumais, T. Letsche, M. Littman, and T. Landauer, \"Automatic cross-language retrieval using latent semantic indexing\", in AAAI Symposium on CrossLanguage Text and Speech Retrieval. 1997, American Association for Artificial Intelligence.",
"links": null
},
"BIBREF1": {
"ref_id": "b1",
"title": "Automatic 3-language cross-language information retrieval with latent semantic indexing",
"authors": [
{
"first": "Bob",
"middle": [],
"last": "Rehder",
"suffix": ""
},
{
"first": "Michael",
"middle": [
"L"
],
"last": "Littman",
"suffix": ""
},
{
"first": "Susan",
"middle": [],
"last": "Dumais",
"suffix": ""
},
{
"first": "Thomas",
"middle": [
"K"
],
"last": "Landauer",
"suffix": ""
}
],
"year": 1997,
"venue": "Sixth Text REtrieval Conference",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Bob Rehder, Michael L. Littman, Susan Dumais, and Thomas K. Landauer, \"Automatic 3-language cross-language information retrieval with latent semantic indexing\", in Sixth Text REtrieval Conference (TREC-6), 1997.",
"links": null
},
"BIBREF2": {
"ref_id": "b2",
"title": "Translingual information retrieval: Learning from bilingual corpora",
"authors": [
{
"first": "Yiming",
"middle": [],
"last": "Yang",
"suffix": ""
},
{
"first": "Jaime",
"middle": [
"G"
],
"last": "Carbonell",
"suffix": ""
},
{
"first": "Ralf",
"middle": [
"D"
],
"last": "Brown",
"suffix": ""
},
{
"first": "Robert",
"middle": [
"E"
],
"last": "Frederking",
"suffix": ""
}
],
"year": 1998,
"venue": "Artificial Intelligence",
"volume": "103",
"issue": "",
"pages": "323--345",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Yiming Yang, Jaime G. Carbonell, Ralf D. Brown, and Robert E. Frederking, \"Translin- gual information retrieval: Learning from bilingual corpora\", Artificial Intelligence, vol. 103, pp. 323-345, 1998.",
"links": null
},
"BIBREF3": {
"ref_id": "b3",
"title": "Cross-language text retrieval based on document translation using japanese-to-korean mt system",
"authors": [
{
"first": "I",
"middle": [
"S"
],
"last": "Oh-Wook Kwon",
"suffix": ""
},
{
"first": "J-H",
"middle": [],
"last": "Kang",
"suffix": ""
},
{
"first": "G",
"middle": [
"B"
],
"last": "Lee",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Lee",
"suffix": ""
}
],
"year": 1997,
"venue": "",
"volume": "",
"issue": "",
"pages": "101--106",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Oh-Wook Kwon, I.S. Kang, J-H Lee, and G.B. Lee, \"Cross-language text retrieval based on document translation using japanese-to-korean mt system\", in NLPRS, 1997, pp. 101- 106.",
"links": null
},
"BIBREF4": {
"ref_id": "b4",
"title": "Document translation for the cross-language text retrieval at the university of maryland",
"authors": [
{
"first": "W",
"middle": [],
"last": "Douglas",
"suffix": ""
},
{
"first": "Paul",
"middle": [],
"last": "Oard",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Hackett",
"suffix": ""
}
],
"year": null,
"venue": "the Sixth Text REtrieval Conference (TREC-6)",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Douglas W. Oard and Paul Hackett, \"Document translation for the cross-language text retrieval at the university of maryland\", in the Sixth Text REtrieval Conference (TREC-6).",
"links": null
},
"BIBREF5": {
"ref_id": "b5",
"title": "Kunlp system for ntcir-4 korean-english cross-language information retrieval",
"authors": [
{
"first": "Hee-Cheol",
"middle": [],
"last": "Seo",
"suffix": ""
},
{
"first": "Sang-Bum",
"middle": [],
"last": "Kim",
"suffix": ""
},
{
"first": "Ho-Gun",
"middle": [],
"last": "Lim",
"suffix": ""
},
{
"first": "Hae-Chang",
"middle": [],
"last": "Rim",
"suffix": ""
}
],
"year": 2004,
"venue": "NTCIR-4",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Hee-Cheol Seo, Sang-Bum Kim, Ho-Gun Lim, and Hae-Chang Rim, \"Kunlp system for ntcir-4 korean-english cross-language information retrieval\", in NTCIR-4, Tokyo, 2004.",
"links": null
},
"BIBREF6": {
"ref_id": "b6",
"title": "Overview of clir task at the sixth ntcir workshop",
"authors": [
{
"first": "Kazuaki",
"middle": [],
"last": "Kishida",
"suffix": ""
},
{
"first": "Sukhoon",
"middle": [],
"last": "Kuang-Hua Chen",
"suffix": ""
},
{
"first": "Kazuko",
"middle": [],
"last": "Lee",
"suffix": ""
},
{
"first": "Noriko",
"middle": [],
"last": "Kuriyama",
"suffix": ""
},
{
"first": "Hsin-Hsi",
"middle": [],
"last": "Kando",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Chen",
"suffix": ""
}
],
"year": 2007,
"venue": "Proceedings of NTCIR-6 Workshop Meeting",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Kazuaki Kishida, Kuang-hua Chen, Sukhoon Lee, Kazuko Kuriyama, Noriko Kando, and Hsin-Hsi Chen, \"Overview of clir task at the sixth ntcir workshop\", Proceedings of NTCIR-6 Workshop Meeting, 2007.",
"links": null
},
"BIBREF7": {
"ref_id": "b7",
"title": "Asqa: Academia sinica question answering system for ntcir-5 qa",
"authors": [
{
"first": "Cheng-Wei",
"middle": [],
"last": "Lee",
"suffix": ""
},
{
"first": "Cheng-Wei",
"middle": [],
"last": "Shih",
"suffix": ""
},
{
"first": "Min-Yuh",
"middle": [],
"last": "Day",
"suffix": ""
},
{
"first": "Tzong-Han",
"middle": [],
"last": "Tsai",
"suffix": ""
},
{
"first": "Tian-Jian",
"middle": [],
"last": "Jiang",
"suffix": ""
},
{
"first": "Chia-Wei",
"middle": [],
"last": "Wu",
"suffix": ""
},
{
"first": "Cheng-Lung",
"middle": [],
"last": "Sung",
"suffix": ""
},
{
"first": "Yu-Ren",
"middle": [],
"last": "Chen",
"suffix": ""
},
{
"first": "Shih-Hung",
"middle": [],
"last": "Wu",
"suffix": ""
},
{
"first": "Wen-Lian",
"middle": [],
"last": "Hsu",
"suffix": ""
}
],
"year": 2005,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Cheng-Wei Lee, Cheng-Wei Shih, Min-Yuh Day, Tzong-Han Tsai, Tian-Jian Jiang, Chia- Wei Wu, Cheng-Lung Sung, Yu-Ren Chen, Shih-Hung Wu, and Wen-Lian Hsu, \"Asqa: Academia sinica question answering system for ntcir-5 qa\", in NTCIR-5, Tokyo, 2005.",
"links": null
},
"BIBREF8": {
"ref_id": "b8",
"title": "A study of information seeking and retrieving",
"authors": [
{
"first": "Tefko",
"middle": [],
"last": "Saracevic",
"suffix": ""
},
{
"first": "Paul",
"middle": [],
"last": "Kantor",
"suffix": ""
},
{
"first": "Alice",
"middle": [
"Y"
],
"last": "Chamis",
"suffix": ""
},
{
"first": "Donna",
"middle": [],
"last": "Trivison",
"suffix": ""
}
],
"year": 1988,
"venue": "Journal of the American Society for Information Science",
"volume": "39",
"issue": "3",
"pages": "161--176",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Tefko Saracevic, Paul Kantor, Alice Y. Chamis, and Donna Trivison, \"A study of infor- mation seeking and retrieving\", Journal of the American Society for Information Science, vol. 39, no. 3, pp. 161-176, 1988.",
"links": null
}
},
"ref_entries": {
"TABREF1": {
"text": "Evaluation Results",
"content": "<table><tr><td>Run</td><td colspan=\"4\">Rigid MAP R-precision MAP R-precision Relax</td></tr><tr><td>KC-title-run</td><td>0.1118</td><td>0.1420</td><td>0.1392</td><td>0.1781</td></tr><tr><td colspan=\"2\">KC-description-run 0.1022</td><td>0.1311</td><td>0.1274</td><td>0.1760</td></tr><tr><td>CC-title-run</td><td>0.1501</td><td>0.1961</td><td>0.2141</td><td>0.2747</td></tr><tr><td colspan=\"2\">CC-description-run 0.1567</td><td>0.2111</td><td>0.2157</td><td>0.2788</td></tr></table>",
"num": null,
"html": null,
"type_str": "table"
}
}
}
} |