File size: 56,773 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 |
{
"paper_id": "I05-1031",
"header": {
"generated_with": "S2ORC 1.0.0",
"date_generated": "2023-01-19T07:24:51.217477Z"
},
"title": "Analysis of an Iterative Algorithm for Term-Based Ontology Alignment",
"authors": [
{
"first": "Shisanu",
"middle": [],
"last": "Tongchim",
"suffix": "",
"affiliation": {
"laboratory": "Thai Computational Linguistics Laboratory",
"institution": "National Institute of Information and Communications Technology",
"location": {
"addrLine": "Paholyothin Road, Klong 1",
"postCode": "112, 12120",
"settlement": "Klong Luang, Pathumthani",
"country": "Thailand"
}
},
"email": "shisanu@tcllab.org"
},
{
"first": "Canasai",
"middle": [],
"last": "Kruengkrai",
"suffix": "",
"affiliation": {
"laboratory": "Thai Computational Linguistics Laboratory",
"institution": "National Institute of Information and Communications Technology",
"location": {
"addrLine": "Paholyothin Road, Klong 1",
"postCode": "112, 12120",
"settlement": "Klong Luang, Pathumthani",
"country": "Thailand"
}
},
"email": "canasai@tcllab.org"
},
{
"first": "Virach",
"middle": [],
"last": "Sornlertlamvanich",
"suffix": "",
"affiliation": {
"laboratory": "Thai Computational Linguistics Laboratory",
"institution": "National Institute of Information and Communications Technology",
"location": {
"addrLine": "Paholyothin Road, Klong 1",
"postCode": "112, 12120",
"settlement": "Klong Luang, Pathumthani",
"country": "Thailand"
}
},
"email": "virach@tcllab.org"
},
{
"first": "Prapass",
"middle": [],
"last": "Srichaivattana",
"suffix": "",
"affiliation": {
"laboratory": "Thai Computational Linguistics Laboratory",
"institution": "National Institute of Information and Communications Technology",
"location": {
"addrLine": "Paholyothin Road, Klong 1",
"postCode": "112, 12120",
"settlement": "Klong Luang, Pathumthani",
"country": "Thailand"
}
},
"email": "prapass@tcllab.org"
},
{
"first": "Hitoshi",
"middle": [],
"last": "Isahara",
"suffix": "",
"affiliation": {
"laboratory": "Thai Computational Linguistics Laboratory",
"institution": "National Institute of Information and Communications Technology",
"location": {
"addrLine": "Paholyothin Road, Klong 1",
"postCode": "112, 12120",
"settlement": "Klong Luang, Pathumthani",
"country": "Thailand"
}
},
"email": "isahara@nict.go.jp"
}
],
"year": "",
"venue": null,
"identifiers": {},
"abstract": "This paper analyzes the results of automatic concept alignment between two ontologies. We use an iterative algorithm to perform concept alignment. The algorithm uses the similarity of shared terms in order to find the most appropriate target concept for a particular source concept. The results show that the proposed algorithm not only finds the relation between the target concepts and the source concepts, but the algorithm also shows some flaws in the ontologies. These results can be used to improve the correctness of the ontologies.",
"pdf_parse": {
"paper_id": "I05-1031",
"_pdf_hash": "",
"abstract": [
{
"text": "This paper analyzes the results of automatic concept alignment between two ontologies. We use an iterative algorithm to perform concept alignment. The algorithm uses the similarity of shared terms in order to find the most appropriate target concept for a particular source concept. The results show that the proposed algorithm not only finds the relation between the target concepts and the source concepts, but the algorithm also shows some flaws in the ontologies. These results can be used to improve the correctness of the ontologies.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Abstract",
"sec_num": null
}
],
"body_text": [
{
"text": "To date, several linguistic ontologies in different languages have been developed independently. The integration of these existing ontologies is useful for many applications. Aligning concepts between ontologies is often done by humans, which is an expensive and time-consuming process. This motivates us to find an automatic method to perform such task. However, the hierarchical structures of ontologies are quite different. The structural inconsistency is a common problem [1] . Developing a practical algorithm that is able to deal with this problem is a challenging issue.",
"cite_spans": [
{
"start": 476,
"end": 479,
"text": "[1]",
"ref_id": "BIBREF0"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "The objective of this research is to investigate an automated technique for ontology alignment. The proposed algorithm links concepts between two ontologies, namely the MMT semantic hierarchy and the EDR concept dictionary. The algorithm finds the most appropriate target concept for a given source concept in the top-down manner. The experimental results show that the algorithm can find reasonable concept mapping between these ontologies. Moreover, the results also suggest that this algorithm is able to detect flaws and inconsistency in the ontologies. These results can be used for developing and improving the ontologies by lexicographers.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "The rest of this paper is organized as follows: Section 2 discusses related work. Section 3 provides the description of the proposed algorithm. Section 4 presents experimental results and discussion. Finally, Section 5 concludes our work.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "Daud\u00e9 et al. [2] used a relaxation labeling algorithm -a constraint satisfaction algorithm -to map the verbal, adjectival and adverbial parts between two different WordNet versions, namely WordNet 1.5 and WordNet 1.6. The structural constraints are used by the algorithm to adjust the weights for the connections between WN1.5 and WN1.6. Later, some non-structural constraints are included in order to improve the performance [3] .",
"cite_spans": [
{
"start": 13,
"end": 16,
"text": "[2]",
"ref_id": "BIBREF1"
},
{
"start": 426,
"end": 429,
"text": "[3]",
"ref_id": "BIBREF2"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Related Work",
"sec_num": "2"
},
{
"text": "Asanoma [4] presented an alignment technique between the noun part of WordNet and Goi-Taikei 's Ontology. The proposed technique utilizes sets of Japanese and/or English words and semantic classes from dictionaries in an MT system, namely ALT-J/E. Chen and Fung [5] proposed an automatic technique to associate the English FrameNet lexical entries to the appropriate Chinese word senses. Each FrameNet lexical entry is linked to Chinese word senses of a Chinese ontology database called HowNet. In the beginning, each FrameNet lexical entry is associated with Chinese word senses whose part-of-speech is the same and Chinese word/phrase is one of the translations. In the second stage of the algorithm, some links are pruned out by analyzing contextual lexical entries from the same semantic frame. In the last stage, some pruned links are recovered if their scores are greater than the calculated threshold value.",
"cite_spans": [
{
"start": 8,
"end": 11,
"text": "[4]",
"ref_id": "BIBREF3"
},
{
"start": 262,
"end": 265,
"text": "[5]",
"ref_id": "BIBREF4"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Related Work",
"sec_num": "2"
},
{
"text": "Ngai et al. [6] also conducted some experiments by using HowNet. They presented a method for performing alignment between HowNet and WordNet. They used a word-vector based method which was adopted from techniques used in machine translation and information retrieval. Recently, Yeh et al. [7] constructed a bilingual ontology by aligning Chinese words in HowNet with corresponding synsets defined in WordNet. Their alignment approach utilized the co-occurrence of words in a parallel bilingual corpus.",
"cite_spans": [
{
"start": 12,
"end": 15,
"text": "[6]",
"ref_id": "BIBREF5"
},
{
"start": 289,
"end": 292,
"text": "[7]",
"ref_id": "BIBREF6"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Related Work",
"sec_num": "2"
},
{
"text": "Khan and Hovy [8] presented an algorithm to combine an Arabic-English dictionary with WordNet. Their algorithm also tries to find links from Arabic words to WordNet first. Then, the algorithm prunes out some links by trying to find a generalization concept.",
"cite_spans": [
{
"start": 14,
"end": 17,
"text": "[8]",
"ref_id": "BIBREF7"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Related Work",
"sec_num": "2"
},
{
"text": "Doan et al. [9] proposed a three steps approach for mapping between ontologies on the semantic web. The first step used machine learning techniques to determine the joint distribution of any concept pair. Then, a user-supplied similarity function is used to compute similarity of concept pairs based on the joint distribution from the first step. In the final step, a relaxation labeling algorithm is used to find the mapping configuration based on the similarity from the previous step.",
"cite_spans": [
{
"start": 12,
"end": 15,
"text": "[9]",
"ref_id": "BIBREF8"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Related Work",
"sec_num": "2"
},
{
"text": "In this section, we describe an approach for ontology alignment based on term distribution. To alleviate the structural computation problem, we assume that the considered ontology structure has only the hierarchical (or taxonomic) relation. One may simply think of this ontology structure as a general tree, where each node of the tree is equivalent to a concept.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Proposed Algorithm",
"sec_num": "3"
},
{
"text": "Given two ontologies called the source ontology T s and the target ontology T t , our objective is to align all concepts (or semantic classes) between these two ontologies. Each ontology consists of the concepts, denoted by C 1 , . . . , C k . In general, the concepts and their corresponding relations of each ontology can be significantly different due to the theoretical background used in the construction process. However, for the lexical ontologies such as the MMT semantic hierarchy and the EDR concept dictionary, it is possible that the concepts may contain shared members in terms of English words. Thus, we can match the concepts between two ontologies using the similarity of the shared words.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Proposed Algorithm",
"sec_num": "3"
},
{
"text": "In order to compute the similarity between two concepts, we must also consider their related child concepts. Given a root concept C i , if we flatten the hierarchy starting from C i , we obtain a nested cluster, whose largest cluster dominates all sub-clusters. As a result, we can represent the nested cluster with a feature vector c i = (w 1 , . . . , w |V| ) T , where features are the set of unique English words V extracted from both ontologies, and w j is the number of the word j occurring the nested cluster i. We note that a word can occur more than once, since it may be placed in several concepts on the lexical ontology according to its sense.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Proposed Algorithm",
"sec_num": "3"
},
{
"text": "After concepts are represented with the feature vectors, the similarity between any two concepts can be easily computed. A variety of standard similarity measures exists, such as the Dice coefficient, the Jaccard coefficient, and the cosine similarity [10] . In our work, we require a similarity measure that can reflect the degree of the overlap between two concepts. Thus, the Jaccard coefficient is suitable for our task. Recently, Strehl and Ghosh [11] have proposed a version of the Jaccard coefficient called the extended Jaccard similarity that can work with continuous or discrete non-negative features. Let x i be the L 2 norm of a given vector x i . The extended Jaccard similarity can be calculated as follows:",
"cite_spans": [
{
"start": 252,
"end": 256,
"text": "[10]",
"ref_id": "BIBREF9"
},
{
"start": 452,
"end": 456,
"text": "[11]",
"ref_id": "BIBREF10"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Proposed Algorithm",
"sec_num": "3"
},
{
"text": "EQUATION",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [
{
"start": 0,
"end": 8,
"text": "EQUATION",
"ref_id": "EQREF",
"raw_str": "JaccardSim(x i , x j ) = x T i x j x i 2 + x j 2 \u2212 x T i x j .",
"eq_num": "( 1 )"
}
],
"section": "Proposed Algorithm",
"sec_num": "3"
},
{
"text": "We now describe an iterative algorithm for term-based ontology alignment. As mentioned earlier, we formulate that the ontology structure is in the form of the general tree. Our algorithm aligns the concepts on the source ontology T s to the concepts on the target ontology T t by performing search and comparison in the top-down manner.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Proposed Algorithm",
"sec_num": "3"
},
{
"text": "Given a concept C i \u2208 T s , the algorithm attempts to find the most appropriate concept B * \u2208 T t , which is located on an arbitrary level of the hierarchy. The algorithm starts by constructing the feature vectors for the current root concept on the level l and its child concepts on the level l + 1. It then calculates the similarity scores between a given source concept and candidate target concepts. If the similarity scores of the child concepts are not greater than the root concept, then the algorithm terminates. Otherwise, it selects a child concept having the maximum score to be the new root concept, and iterates the same searching procedure. Algorithms 1 and 2 outline our ontology alignment process. Figure 1 shows a simple example that describes how the algorithm works. It begins with finding the most appropriate concept on T t for the root concept 1 \u2208 T s . By flattening the hierarchy starting from given concepts ('1' on T s , and 'a', 'a-b', 'a-c' for T t ), we can represent them with the feature vectors and measure their similarities. On the first iteration, the child concept 'a-c' obtains the maximum score, so it becomes the new root concept. Since the algorithm cannot find improvement on any child concepts in the second iteration, it stops the loop and the target concept 'a-c' is aligned with the source concept '1'. The algorithm proceeds with the same steps by finding the most appropriate concepts on T t for the concepts '1-1' and '1-2'. It finally obtains the resulting concepts 'a-c-f' and 'a-c-g', respectively. ",
"cite_spans": [],
"ref_spans": [
{
"start": 714,
"end": 722,
"text": "Figure 1",
"ref_id": "FIGREF0"
}
],
"eq_spans": [],
"section": "Proposed Algorithm",
"sec_num": "3"
},
{
"text": "Two dictionaries are used in our experiments. The first one is the EDR Electronic Dictionary [12] . The second one is the electronic dictionary of Multilingual Machine Translation (MMT) project [13] .",
"cite_spans": [
{
"start": 93,
"end": 97,
"text": "[12]",
"ref_id": "BIBREF11"
},
{
"start": 194,
"end": 198,
"text": "[13]",
"ref_id": "BIBREF12"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Data Sets",
"sec_num": "4.1"
},
{
"text": "The EDR Electronic Dictionary consists of lexical knowledge of Japanese and English divided into several sub-dictionaries (e.g., the word dictionary, the bilingual dictionary, the concept dictionary, and the co-occurrence dictionary) and the EDR corpus. In the revised version (version 1.5), the Japanese word dictionary contains 250,000 words, while the English word dictionary contains 190,000 words. The concept dictionary holds information on the 400,000 concepts that are listed in the word dictionary. Each concept is marked with a unique hexadecimal number.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Data Sets",
"sec_num": "4.1"
},
{
"text": "For the MMT dictionary, we use the Thai-English Bilingual Dictionary that contains around 60,000 lexical entries. The Thai-English Bilingual Dictionary also contains semantic information about the case relations and the word concepts. The word concepts are organized in a manner of semantic hierarchy. Each word concept is a group of lexical entries classified and ordered in a hierarchical level of meanings. The MMT semantic hierarchy is composed of 160 concepts.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Data Sets",
"sec_num": "4.1"
},
{
"text": "In our experiments, we used a portion of the MMT semantic hierarchy and the EDR concept dictionary as the source and the target ontologies, respectively. We considered the 'animal' concept as the root concepts and extracted its related concepts. In the EDR concept dictionary, however, the relations among concepts are very complex and organized in the form of the semantic network. Thus, we pruned some links to transform the network to a tree structure. Starting from the 'animal' concept, there are more than 200 sub-concepts (containing about 7,600 words) in the EDR concept dictionary, and 14 sub-concepts (containing about 400 words) in the MMT semantic hierarchy. It is important to note that these two ontologies are considerably different in terms of the number of concepts and words.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Data Sets",
"sec_num": "4.1"
},
{
"text": "The proposed algorithm is used to find appropriate EDR concepts for each one of 14 MMT concepts. The results are shown in Table 1 . From the table, there are 6 relations (marked with the symbol '*') that are manually classified as exact mapping. This classification is done by inspecting the structures of both ontologies by hand. If the definition of a given MMT concept appears in the EDR concept and the algorithm seems to correctly match the most suitable EDR concept, this mapping will be classified as exact mapping. The remaining 8 MMT concepts, e.g. 'cold-blood' and 'amphibian', are mapped to closely related EDR concepts, although they are not considered to be exact mapping. The EDR concepts found by our algorithm for these 8 MMT concepts are considered to be only the subset of the source concepts. For example, the 'amphibian' concept of the MMT is mapped to the 'toad' concept of the EDR. The analysis in the later section will explain why some MMT concepts are mapped to specific sub-concepts.",
"cite_spans": [],
"ref_spans": [
{
"start": 122,
"end": 129,
"text": "Table 1",
"ref_id": "TABREF0"
}
],
"eq_spans": [],
"section": "Experimental Results",
"sec_num": "4.2"
},
{
"text": "Our algorithm works by flattening the hierarchy starting from the considered concept in order to construct a word list represented that concept. The word lists are then compared to match the concepts. In practice, only a portion of word list is intersected. Figure 2 illustrates what happens in general. Note that the EDR concept dictionary is much larger than the MMT semantic hierarchy. Thus, it always has EDR words that are not matched with any MMT words. These words are located in the section 3 of the figure 2. The words in the section 1 are more important since they affects the performance of the algorithm. We assume that the EDR is much larger than the MMT. Therefore, most MMT words should be found in the EDR. The MMT words that cannot found any related EDR words may be results of incorrect spellings, specific words (i.e. only found in Thai language). In case of incorrect spelling and other similar problems, the results of the algorithm can be used to improve the MMT ontology.",
"cite_spans": [],
"ref_spans": [
{
"start": 258,
"end": 266,
"text": "Figure 2",
"ref_id": "FIGREF1"
}
],
"eq_spans": [],
"section": "Experimental Results",
"sec_num": "4.2"
},
{
"text": "By analyzing the results, we can classify the MMT words that cannot find any associated EDR words into 4 categories.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Experimental Results",
"sec_num": "4.2"
},
{
"text": "Incorrect spelling or wrong grammar : Some English words in the MMT semantic hierarchy are simply incorrect spelling, or they are written with wrong grammar. For example, one description of a tiger species is written as 'KIND A TIGER'. Actually, this instance should be 'KIND OF A TIGER'. The algorithm can be used to find words that possible have such a problem. Then, the words can be corrected by lexicographers. 2. Inconsistency : The English translation of Thai words in the MMT semantic hierarchy was performed by several lexicographers. When dealing with Thai words that do not have exact English words, lexicographers usually enter phrases as descriptions of these words. Since there is no standard of writing the descriptions, these is incompatibility between descriptions that explain the same concept. For example, the following phrases are used to describe fishes that their English names are not known.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "1.",
"sec_num": null
},
{
"text": "-Species of fish -A kind of fish -Species of fresh water fish 3. Thai specific words : The words that we used in our experiments are animals. Several animals are region specific species. Therefore, they may not have any associated English words. In this case, some words are translated by using short phrases as English descriptions of these Thai words. Another way to translate these words is to use scientific names of species.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "1.",
"sec_num": null
},
{
"text": "The problems mentioned earlier make it more difficult to match concepts by the algorithm. However, we can use the algorithm to identify where the problems occur. Then, we can use these results to improve the MMT ontology.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "1.",
"sec_num": null
},
{
"text": "The proposed algorithm works in the top-down manner. That is, the algorithm attempts to find the most appropriate concept from the top level, and it will move down if the lower concepts yield better scores. In order to analyze the algorithm, we trace the algorithm during moving through the EDR concepts. The first example of the bird concept alignment is shown in Table 2 . The concept alignment of this example is considered to be exact mapping. The first column indicates the level of EDR concepts. The second and third columns indicate the number of MMT words and the number of EDR words after flattening respectively. The fourth column shows the number of intersected words between the MMT and the EDR. From the table, the algorithm moves through the EDR concepts in order to find the most specific concept that still maintains shared terms. This example shows that the algorithm passes through 3 concepts until it stops at the 'bird' concept of the EDR. At the final step, the algorithm decides to trade few shared terms for a more specific EDR concept. Note that the MMT is not completely cleaned. When moving down to the EDR bird concept, three shared terms are lost. Our analysis shows that these terms are bat species. They are all wrongly classified to the MMT bird concept by some lexicographers. Thus, these shared terms will not intersect with any words in the EDR bird concept when the algorithm proceeds to the lower step. This result suggests that our algorithm is quite robust. The algorithm still finds an appropriate concept even the MMT ontology has some flaws.",
"cite_spans": [],
"ref_spans": [
{
"start": 365,
"end": 372,
"text": "Table 2",
"ref_id": "TABREF2"
}
],
"eq_spans": [],
"section": "1.",
"sec_num": null
},
{
"text": "Another analysis of exact mapping is shown in Table 3 . The algorithm moves through 4 concepts until matching the EDR snake concept with the MMT snake concept. In this example, the number of members in the MMT snake concept is quite small. However, the number of shared terms is sufficient to correctly locate the EDR snake concept. The third example shown in Table 4 illustrates the case that is considered to be subset mapping. That is, the EDR concept selected by the algorithm is subconcept of the MMT concept. This case happens several times since the EDR is more fine-grained than the MMT. If the members of MMT concept do not cover enough, the algorithm tends to return only sub-concepts. From the table, the MMT amphibian concept covers only toad and frog species (3 members). Thus, the algorithm moves down to a very specific concept, namely the EDR toad concept. Another example of subset mapping is shown in Table 5 . This example also shows that the members of MMT concept do not cover enough. These results can be used to improve the MMT ontology. If the MMT concepts are extended enough, we expect that the correctness of alignment should be improved. Intersected words 1 3 2112 2 2 3 1288 2 3 3 23 2 4 3 16 2 5 3 2 1 ",
"cite_spans": [],
"ref_spans": [
{
"start": 46,
"end": 53,
"text": "Table 3",
"ref_id": "TABREF3"
},
{
"start": 360,
"end": 367,
"text": "Table 4",
"ref_id": "TABREF4"
},
{
"start": 919,
"end": 926,
"text": "Table 5",
"ref_id": "TABREF5"
},
{
"start": 1166,
"end": 1242,
"text": "Intersected words 1 3 2112 2 2 3 1288 2 3 3 23 2 4 3 16 2 5",
"ref_id": "TABREF0"
}
],
"eq_spans": [],
"section": "1.",
"sec_num": null
},
{
"text": "We have proposed an iterative algorithm to deal with the problem of automated ontology alignment. This algorithm works in the top-down manner by using the similarity of the terms from each ontology. We use two dictionaries in our experiment, namely the MMT semantic hierarchy and the EDR concept dictionary.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Conclusion",
"sec_num": "5"
},
{
"text": "The results show that the algorithm can find reasonable EDR concepts for given MMT concepts. Moreover, the results also suggest that the algorithm can be used as a tool to locate flaws in the MMT ontology. These results can be used to improve the ontology. There are several possible extensions to this study. The first one is to examine this algorithm with larger data sets or other ontologies. The second one is to improve and correct the ontologies by using the results from the algorithm. Then, we plan to apply this algorithm to the corrected ontologies, and examine the correctness of the results. The third one is to use structural information of ontologies in order to improve the correctness.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Conclusion",
"sec_num": "5"
}
],
"back_matter": [],
"bib_entries": {
"BIBREF0": {
"ref_id": "b0",
"title": "Machine Readable Dictionaries: What have we learned, where do we go",
"authors": [
{
"first": "N",
"middle": [],
"last": "Ide",
"suffix": ""
},
{
"first": "J",
"middle": [],
"last": "V\u00e9ronis",
"suffix": ""
}
],
"year": 1994,
"venue": "Proceedings of the International Workshop on the Future of Lexical Research",
"volume": "",
"issue": "",
"pages": "137--146",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Ide, N. and V\u00e9ronis, J.: Machine Readable Dictionaries: What have we learned, where do we go?. Proceedings of the International Workshop on the Future of Lexical Research, Beijing, China (1994) 137-146",
"links": null
},
"BIBREF1": {
"ref_id": "b1",
"title": "Mapping WordNets Using Structural Information",
"authors": [
{
"first": "J",
"middle": [],
"last": "Daud\u00e9",
"suffix": ""
},
{
"first": "L",
"middle": [],
"last": "Padr\u00f3",
"suffix": ""
},
{
"first": "G",
"middle": [],
"last": "Rigau",
"suffix": ""
}
],
"year": 2000,
"venue": "Proceedings of the 38th Annual Meeting of the Association for Computational Linguistics",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Daud\u00e9, J., Padr\u00f3, L. and Rigau, G.: Mapping WordNets Using Structural Informa- tion. Proceedings of the 38th Annual Meeting of the Association for Computational Linguistics, Hong Kong, (2000)",
"links": null
},
"BIBREF2": {
"ref_id": "b2",
"title": "A Complete WN1.5 to WN1.6 Mapping. Proceedings of NAACL Workshop \"WordNet and Other Lexical Resources: Applications, Extensions and Customizations",
"authors": [
{
"first": "J",
"middle": [],
"last": "Daud\u00e9",
"suffix": ""
},
{
"first": "L",
"middle": [],
"last": "Padr\u00f3",
"suffix": ""
},
{
"first": "G",
"middle": [],
"last": "Rigau",
"suffix": ""
}
],
"year": 2001,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Daud\u00e9, J., Padr\u00f3, L. and Rigau, G.: A Complete WN1.5 to WN1.6 Mapping. Proceedings of NAACL Workshop \"WordNet and Other Lexical Resources: Appli- cations, Extensions and Customizations\", Pittsburg, PA, United States, (2001)",
"links": null
},
"BIBREF3": {
"ref_id": "b3",
"title": "Alignment of Ontologies: WordNet and Goi-Taikei",
"authors": [
{
"first": "N",
"middle": [],
"last": "Asanoma",
"suffix": ""
}
],
"year": 2001,
"venue": "Proceedings of NAACL Workshop \"WordNet and Other Lexical Resources: Applications, Extensions and Customizations",
"volume": "",
"issue": "",
"pages": "89--94",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Asanoma, N.: Alignment of Ontologies: WordNet and Goi-Taikei. Proceedings of NAACL Workshop \"WordNet and Other Lexical Resources: Applications, Exten- sions and Customizations\", Pittsburg, PA, United States, (2001) 89-94",
"links": null
},
"BIBREF4": {
"ref_id": "b4",
"title": "Automatic Construction of an English-Chinese Bilingual FrameNet",
"authors": [
{
"first": "B",
"middle": [],
"last": "Chen",
"suffix": ""
},
{
"first": "P",
"middle": [],
"last": "Fung",
"suffix": ""
}
],
"year": 2004,
"venue": "Proceedings of Human Language Technology conference",
"volume": "",
"issue": "",
"pages": "29--32",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Chen, B. and Fung, P.: Automatic Construction of an English-Chinese Bilingual FrameNet. Proceedings of Human Language Technology conference, Boston, MA (2004) 29-32",
"links": null
},
"BIBREF5": {
"ref_id": "b5",
"title": "Identifying Concepts Across Languages: A First Step towards a Corpus-based Approach to Automatic Ontology Alignment",
"authors": [
{
"first": "G",
"middle": [],
"last": "Ngai",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Carpuat",
"suffix": ""
},
{
"first": "P",
"middle": [],
"last": "Fung",
"suffix": ""
}
],
"year": 2002,
"venue": "Proceedings of the 19th International Conference on Computational Linguistics",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Ngai, G., Carpuat , M. and Fung, P.: Identifying Concepts Across Languages: A First Step towards a Corpus-based Approach to Automatic Ontology Alignment. Proceedings of the 19th International Conference on Computational Linguistics, Taipei, Taiwan (2002)",
"links": null
},
"BIBREF6": {
"ref_id": "b6",
"title": "Automated Alignment and Extraction of a Bilingual Ontology for Cross-Language Domain-Specific Applications",
"authors": [
{
"first": "J.-F",
"middle": [],
"last": "Yeh",
"suffix": ""
},
{
"first": "C.-H",
"middle": [],
"last": "Wu",
"suffix": ""
},
{
"first": "M.-J",
"middle": [],
"last": "Chen",
"suffix": ""
},
{
"first": "L.-C",
"middle": [],
"last": "Yu",
"suffix": ""
}
],
"year": 2005,
"venue": "International Journal of Computational Linguistics and Chinese Language Processing",
"volume": "10",
"issue": "",
"pages": "35--52",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Yeh, J.-F., Wu, C.-H., Chen, M.-J. and Yu, L.-C.: Automated Alignment and Extraction of a Bilingual Ontology for Cross-Language Domain-Specific Applica- tions. International Journal of Computational Linguistics and Chinese Language Processing. 10 (2005) 35-52",
"links": null
},
"BIBREF7": {
"ref_id": "b7",
"title": "Improving the Precision of Lexicon-to-Ontology Alignment Algorithms",
"authors": [
{
"first": "L",
"middle": [],
"last": "Khan",
"suffix": ""
},
{
"first": "E",
"middle": [],
"last": "Hovy",
"suffix": ""
}
],
"year": 1997,
"venue": "Proceedings of AMTA/SIG-IL First Workshop on Interlinguas",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Khan, L. and Hovy, E.: Improving the Precision of Lexicon-to-Ontology Alignment Algorithms. Proceedings of AMTA/SIG-IL First Workshop on Interlinguas, San Diego, CA (1997)",
"links": null
},
"BIBREF8": {
"ref_id": "b8",
"title": "Learning to Map Between Ontologies on the Semantic Web",
"authors": [
{
"first": "A",
"middle": [],
"last": "Doan",
"suffix": ""
},
{
"first": "J",
"middle": [],
"last": "Madhavan",
"suffix": ""
},
{
"first": "P",
"middle": [],
"last": "Domingos",
"suffix": ""
},
{
"first": "A",
"middle": [],
"last": "Halevy",
"suffix": ""
}
],
"year": 2002,
"venue": "Proceedings of the 11th international conference on World Wide Web",
"volume": "",
"issue": "",
"pages": "662--673",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Doan, A., Madhavan, J., Domingos, P., and Halevy, A.: Learning to Map Between Ontologies on the Semantic Web. Proceedings of the 11th international conference on World Wide Web, ACM Press (2002) 662-673",
"links": null
},
"BIBREF9": {
"ref_id": "b9",
"title": "Foundations of Statistical Natural Language Processing",
"authors": [
{
"first": "C",
"middle": [
"D"
],
"last": "Manning",
"suffix": ""
},
{
"first": "H",
"middle": [],
"last": "Sch\u00fctze",
"suffix": ""
}
],
"year": 1999,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Manning, C. D., and Sch\u00fctze, H.: Foundations of Statistical Natural Language Processing. MIT Press. Cambridge, MA (1999)",
"links": null
},
"BIBREF10": {
"ref_id": "b10",
"title": "Impact of Similarity Measures on Webpage Clustering",
"authors": [
{
"first": "A",
"middle": [],
"last": "Strehl",
"suffix": ""
},
{
"first": "J",
"middle": [],
"last": "Ghosh",
"suffix": ""
},
{
"first": "R",
"middle": [
"J"
],
"last": "Mooney",
"suffix": ""
}
],
"year": 2000,
"venue": "Proceedings of AAAI Workshop on AI for Web Search",
"volume": "",
"issue": "",
"pages": "58--64",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Strehl, A., Ghosh, J., and Mooney, R. J.: Impact of Similarity Measures on Web- page Clustering. Proceedings of AAAI Workshop on AI for Web Search (2000) 58-64",
"links": null
},
"BIBREF11": {
"ref_id": "b11",
"title": "An Overview of the EDR Electronic Dictionary and the Current Status of Its Utilization",
"authors": [
{
"first": "H",
"middle": [],
"last": "Miyoshi",
"suffix": ""
},
{
"first": "K",
"middle": [],
"last": "Sugiyama",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Kobayashi",
"suffix": ""
},
{
"first": "T",
"middle": [],
"last": "Ogino",
"suffix": ""
}
],
"year": 1996,
"venue": "Proceedings of the 16th International Conference on Computational Linguistics",
"volume": "",
"issue": "",
"pages": "1090--1093",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Miyoshi, H., Sugiyama, K., Kobayashi, M. and Ogino, T.: An Overview of the EDR Electronic Dictionary and the Current Status of Its Utilization. Proceedings of the 16th International Conference on Computational Linguistics (1996) 1090-1093",
"links": null
},
"BIBREF12": {
"ref_id": "b12",
"title": "CICC: Thai Basic Dictionary. Center of the International Cooperation for Computerization",
"authors": [],
"year": 1995,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "CICC: Thai Basic Dictionary. Center of the International Cooperation for Com- puterization, Technical Report 6-CICC-MT55 (1995)",
"links": null
}
},
"ref_entries": {
"FIGREF0": {
"uris": null,
"num": null,
"type_str": "figure",
"text": "An example of finding the most appropriate concept on Tt for the root concept 1 \u2208 Ts"
},
"FIGREF1": {
"uris": null,
"num": null,
"type_str": "figure",
"text": "A schematic of aligned concepts"
},
"TABREF0": {
"num": null,
"text": "OntologyAlignmentinput: The source ontology Ts and the target ontology Tt. output : The set of the aligned concepts A.",
"html": null,
"type_str": "table",
"content": "<table><tr><td>begin</td></tr><tr><td>Set the starting level, l \u2190 0;</td></tr><tr><td>while Ts l \u2264 Ts max do</td></tr><tr><td>Find all child concepts on this level, {Ci} k i=1 \u2208 Ts l ;</td></tr><tr><td>Flatten {Ci} k i=1 and build their corresponding feature vectors, {ci} k i=1 ;</td></tr><tr><td>For each ci, find the best matched concepts on Tt,</td></tr><tr><td>B \u2190 FindBestMatched(ci);</td></tr><tr><td>A \u2190 A \u222a {B, Ci};</td></tr><tr><td>Set l \u2190 l + 1;</td></tr><tr><td>end</td></tr><tr><td>end</td></tr><tr><td>Algorithm 2. FindBestMatched(c i )</td></tr><tr><td>begin</td></tr><tr><td>Set the starting level, l \u2190 0;</td></tr><tr><td>BestConcept \u2190 Tt(root concept);</td></tr><tr><td>repeat</td></tr><tr><td>stmp \u2190 JaccardSim(ci, BestConcept);</td></tr><tr><td>if Tt l > Tt max then</td></tr><tr><td>return BestConcept;</td></tr><tr><td>Find all child concepts on this level, {B} h j=1 \u2208 Tt l ;</td></tr><tr><td>Flatten {Bj } h j=1 and build corresponding feature vectors, {bj } h i=1 ;</td></tr><tr><td>sj * \u2190 argmax j JaccardSim(ci, {bj } h j=1 );</td></tr><tr><td>if sj * > stmp then</td></tr><tr><td>BestConcept \u2190 Bj * ;</td></tr><tr><td>Set l \u2190 l + 1;</td></tr><tr><td>until BestConcept does not change;</td></tr><tr><td>return BestConcept;</td></tr><tr><td>end</td></tr></table>"
},
"TABREF1": {
"num": null,
"text": "Results of aligned concepts between the MMT and the EDR",
"html": null,
"type_str": "table",
"content": "<table><tr><td/><td>MMT concept</td><td>EDR concept</td></tr><tr><td colspan=\"2\">vertebrate</td><td>vertebrate *</td></tr><tr><td colspan=\"2\">| \u2192 warm-blood</td><td>mammal</td></tr><tr><td>|</td><td>|\u2192 mammal</td><td>mammal *</td></tr><tr><td>|</td><td>|\u2192 bird</td><td>bird *</td></tr><tr><td>|</td><td/><td/></tr><tr><td colspan=\"2\">| \u2192 cold-blood</td><td>reptile</td></tr><tr><td/><td>| \u2192 fish</td><td>fish *</td></tr><tr><td/><td>| \u2192 amphibian</td><td>toad</td></tr><tr><td/><td>| \u2192 reptile</td><td>reptile *</td></tr><tr><td/><td>| \u2192 snake</td><td>snake *</td></tr><tr><td colspan=\"2\">invertebrate</td><td>squid</td></tr><tr><td colspan=\"2\">| \u2192 worm</td><td>leech</td></tr><tr><td colspan=\"2\">| \u2192 insect</td><td>hornet</td></tr><tr><td colspan=\"2\">| \u2192 shellfish</td><td>crab</td></tr><tr><td colspan=\"3\">| \u2192 other sea creature squid</td></tr><tr><td/><td/><td/></tr></table>"
},
"TABREF2": {
"num": null,
"text": "Concept alignment for the 'bird' concept",
"html": null,
"type_str": "table",
"content": "<table><tr><td colspan=\"4\">Level MMT words EDR words Intersected words</td></tr><tr><td>1</td><td>67</td><td>2112</td><td>26</td></tr><tr><td>2</td><td>67</td><td>1288</td><td>26</td></tr><tr><td>3</td><td>67</td><td>373</td><td>23</td></tr></table>"
},
"TABREF3": {
"num": null,
"text": "Concept alignment for the 'snake' concept",
"html": null,
"type_str": "table",
"content": "<table><tr><td colspan=\"4\">Level MMT words EDR words Intersected words</td></tr><tr><td>1</td><td>17</td><td>2112</td><td>8</td></tr><tr><td>2</td><td>17</td><td>1288</td><td>8</td></tr><tr><td>3</td><td>17</td><td>71</td><td>8</td></tr><tr><td>4</td><td>17</td><td>26</td><td>8</td></tr></table>"
},
"TABREF4": {
"num": null,
"text": "Concept alignment for the 'amphibian' concept Level MMT words EDR words",
"html": null,
"type_str": "table",
"content": "<table/>"
},
"TABREF5": {
"num": null,
"text": "Concept alignment for the 'other sea creature' concept",
"html": null,
"type_str": "table",
"content": "<table><tr><td colspan=\"4\">Level MMT words EDR words Intersected words</td></tr><tr><td>1</td><td>17</td><td>2112</td><td>5</td></tr><tr><td>2</td><td>17</td><td>746</td><td>5</td></tr><tr><td>3</td><td>17</td><td>78</td><td>3</td></tr><tr><td>4</td><td>17</td><td>3</td><td>2</td></tr></table>"
}
}
}
} |