File size: 77,505 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 |
{
"paper_id": "O12-3002",
"header": {
"generated_with": "S2ORC 1.0.0",
"date_generated": "2023-01-19T08:03:12.711970Z"
},
"title": "Predicting the Semantic Orientation of Terms in E-HowNet",
"authors": [
{
"first": "",
"middle": [],
"last": "\uf9e1\u653f\u5112",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "National Taiwan University",
"location": {}
},
"email": ""
},
{
"first": "Ru",
"middle": [],
"last": "Li",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "National Taiwan University",
"location": {}
},
"email": ""
},
{
"first": "Chi-Hsin",
"middle": [],
"last": "Yu",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "National Taiwan University",
"location": {}
},
"email": ""
},
{
"first": "Hsin-Hsi",
"middle": [],
"last": "Chen",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "National Taiwan University",
"location": {}
},
"email": "hhchen@ntu.edu.tw"
}
],
"year": "",
"venue": null,
"identifiers": {},
"abstract": "The semantic orientation of terms is fundamental for sentiment analysis in sentence and document levels. Although some Chinese sentiment dictionaries are available, how to predict the orientation of terms automatically is still important. In this paper, we predict the semantic orientation of terms of E-HowNet. We extract many useful features from different sources to represent a Chinese term in E-HowNet, and use a supervised machine learning algorithm to predict its orientation. Our experimental results showed that the proposed approach can achieve 92.33% accuracy.",
"pdf_parse": {
"paper_id": "O12-3002",
"_pdf_hash": "",
"abstract": [
{
"text": "The semantic orientation of terms is fundamental for sentiment analysis in sentence and document levels. Although some Chinese sentiment dictionaries are available, how to predict the orientation of terms automatically is still important. In this paper, we predict the semantic orientation of terms of E-HowNet. We extract many useful features from different sources to represent a Chinese term in E-HowNet, and use a supervised machine learning algorithm to predict its orientation. Our experimental results showed that the proposed approach can achieve 92.33% accuracy.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Abstract",
"sec_num": null
}
],
"body_text": [
{
"text": "\u60c5\u7dd2\u5206\u6790(Sentiment Analysis)\u5728\u73fe\u4eca\u7684\u7db2\uf937\u4e16\u754c\u4e2d\uff0c\u6709\u8a31\u591a\u5be6\u969b\u4e14\u91cd\u8981\u7684\u904b\u7528\uff0c\uf9b5\u5982 \u5f9e\u7db2\uf937\u7684\u8a55\uf941\u6587\u7ae0\u4e2d\u5206\u6790\u6d88\u8cbb\u8005\u5c0d\u7522\u54c1\u7684\u8a55\u50f9\uff0c\u6216\u5206\u6790\u6d88\u8cbb\u8005\u5c0d\u7522\u54c1\u6027\u80fd\u7684\u95dc\u6ce8\u7126\u9ede\u7b49 \u7b49\u3002\uf967\u7ba1\u5c0d\uf906\u5b50\u6216\u6587\u4ef6\u5c64\u6b21\u7684\u60c5\u7dd2\u5206\u6790\uff0c\u610f\ufa0a\u8a5e\u8a5e\u5178\u90fd\u662f\u4e00\u500b\u91cd\u8981\u7684\u8cc7\u6e90\u3002\u901a\u5e38\u610f\ufa0a\u8a5e \u8a5e\u5178\u662f\u7528\u4eba\u5de5\uf92d\u6536\u96c6\u8a5e\u5f59\uff0c\u4e26\u7528\u4eba\u5de5\u6a19\u8a18\u8a5e\u5f59\u7684\u5404\u7a2e\u60c5\u7dd2\u5c6c\u6027\uff0c\u5305\u62ec\u4e3b\u5ba2\u89c0 (subjective or objective)\u3001\u6975\u6027(orientation/polarity)\u3001)\u53ca\u6975\u6027\u7684\u5f37\ufa01(strength) (Esuli & Sebastiani, 2005) \u3002\u9019\u4e9b\u60c5\u7dd2\u5c6c\u6027\u5c0d\uf967\u540c\u7684\u61c9\u7528\u6709\uf967\u540c\u7684\u91cd\u8981\u6027\uff0c\u6a19\u8a18\u96e3\ufa01\u4e5f\u5404\uf967\u76f8\u540c\uff0c\u901a\u5e38\u8a5e\u5f59\u7684 \u6975\u6027\u662f\u6700\u5bb9\uf9e0\u9032\ufa08\u6a19\u8a18\u7684\u5c6c\u6027\u3002 \u6a19\u8a18\u60c5\u7dd2\u5c6c\u6027\u6642\uff0c\u7814\u7a76\u8005\u53ef\u4ee5\u5f9e\uf9b2\u958b\u59cb\u6536\u96c6\u8a5e\u5f59\u4ee5\u5efa\uf9f7\u610f\ufa0a\u8a5e\u8a5e\u5178\uff0c\u5982\u53f0\u5927\u610f\ufa0a\u8a5e \u8a5e\u5178 NTUSD (Ku & Chen, 2007) \u3002\u5728\u53e6\u4e00\u65b9\u9762\uff0c\u4e5f\u6709\u7814\u7a76\u8005\u5617\u8a66\u70ba\u81ea\u7136\u8a9e\u8a00\u8655\uf9e4\u4e2d\u7684\u8a31 \u591a\u73fe\u5b58\u7684\u8cc7\u6e90\uff0c\u6dfb\u52a0\u60c5\u7dd2\u5c6c\u6027\uff0c\u5982 SentiWordNet (Esuli & Sebastiani, 2006a) (Esuli & Sebastiani, 2006b; Kamps, Marx, Mokken, & De Rijke, 2004; Turney & Littman, 2003) Yuen et al.(2004) (Han, Mo, Zuo, & Duan, 2010; Li, Ma, & Guo, 2009; Lu, Song, Zhang, & Tsou, 2010; Yao, Wu, Liu, & Zheng, 2006) (Dietterich, 1998) ",
"cite_spans": [
{
"start": 230,
"end": 256,
"text": "(Esuli & Sebastiani, 2005)",
"ref_id": "BIBREF4"
},
{
"start": 355,
"end": 372,
"text": "(Ku & Chen, 2007)",
"ref_id": "BIBREF11"
},
{
"start": 427,
"end": 454,
"text": "(Esuli & Sebastiani, 2006a)",
"ref_id": "BIBREF5"
},
{
"start": 455,
"end": 482,
"text": "(Esuli & Sebastiani, 2006b;",
"ref_id": "BIBREF6"
},
{
"start": 483,
"end": 521,
"text": "Kamps, Marx, Mokken, & De Rijke, 2004;",
"ref_id": "BIBREF10"
},
{
"start": 522,
"end": 545,
"text": "Turney & Littman, 2003)",
"ref_id": "BIBREF15"
},
{
"start": 546,
"end": 563,
"text": "Yuen et al.(2004)",
"ref_id": null
},
{
"start": 564,
"end": 592,
"text": "(Han, Mo, Zuo, & Duan, 2010;",
"ref_id": "BIBREF7"
},
{
"start": 593,
"end": 613,
"text": "Li, Ma, & Guo, 2009;",
"ref_id": "BIBREF12"
},
{
"start": 614,
"end": 644,
"text": "Lu, Song, Zhang, & Tsou, 2010;",
"ref_id": "BIBREF14"
},
{
"start": 645,
"end": 673,
"text": "Yao, Wu, Liu, & Zheng, 2006)",
"ref_id": "BIBREF16"
},
{
"start": 674,
"end": 692,
"text": "(Dietterich, 1998)",
"ref_id": "BIBREF2"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "\u7dd2\uf941",
"sec_num": "1."
},
{
"text": "\uff0c\u6548\u80fd\u5f9e 67%\u5230 88%\uf967\u7b49\uff0c\u4f46\u56e0\u70ba\u9019\u4e9b\u6f14\u7b97\u6cd5\u6240\u7528\u7684\u8cc7\uf9be\u96c6\u4e26\uf967 \u76f8\u540c\uff0c\u5be6\u9a57\u904e\u7a0b\u53ca\u8a55\u4f30\u6a19\u6e96\u4e5f\uf967\u4e00\u6a23\uff0c(\u6709\u7528 Accuracy\u3001Precision\u3001\u6216 F-Measure)\uff0c\u6240 \u4ee5\u6548\u80fd\u6c92\u6709\u8fa6\u6cd5\u76f4\u63a5\u6bd4\u8f03\u3002 \u5716 1.\u300c\u6c7d\u6cb9\u300d\u7684\u5ee3\u7fa9\u77e5\u7db2\u5b9a\u7fa9\u5f0f \u5728 \u4e2d \u6587 \u7684 \u60c5 \u7dd2 \u5c6c \u6027 \u6a19 \u8a18 \u76f8 \u95dc \u7814 \u7a76 \uff0c",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "\u7dd2\uf941",
"sec_num": "1."
},
{
"text": "\uff0c\u6240\u5f97\u5230\u7684\u6548\u80fd\u5728\uf967\u540c\u7684\u6307\u6a19(Accuracy\u3001Precision\u3001\u6216 F-Measure)\u4e0b\uff0c \u5f9e 89%\u5230 96%\uf967\u7b49\u3002\u56e0\u70ba\u57fa\u6e96\uf967\u540c\uff0c\u9019\u4e9b\u6548\u80fd\u4e00\u6a23\u6c92\u6709\u8fa6\u6cd5\u76f4\u63a5\u6bd4\u8f03\uff0c\u4f46\u76f8\u8f03\u65bc\u82f1\u6587\uff0c \u6210\u7e3e\u5247\u660e\u986f\u63d0\u9ad8\u3002 3. \u7279\u5fb5\u62bd\u53d6\u53ca\u6a5f\u5668\u5b78\u7fd2\u6f14\u7b97\u6cd5 \u7531\u65bc\u6211\u5011\u904b\u7528\u76e3\u7763\u5f0f\u6a5f\u5668\u5b78\u7fd2\u6f14\u7b97\u6cd5\uf92d\u8a13\uf996\u4e8c\u5143\u5206\uf9d0\u5668(binary classifier)\uff0c\u6700\u91cd\u8981\u7684 \u554f\u984c\u662f\u70ba\u8a5e\u5f59\u62bd\u53d6\u51fa\u6709\u7528\u7684\u7279\u5fb5\u3002\u5728\u6b64\uf941\u6587\u4e2d\uff0c\u6211\u5011\u5206\u5225\u5f9e E-HowNet \u53ca Google Chinese Web 5-gram \u9019\uf978\u500b\uf92d\u6e90\u62bd\u53d6\uf978\u5927\uf9d0\u7684\u7279\u5fb5\uff0c\u63a5\u8457\u5c07\u9019\uf978\u500b\uf92d\u6e90\u7684\u7279\u5fb5\u7d44\u5408\u8a13\uf996\u5206\uf9d0\u5668\u3002 \u6b64\u5916\uff0c\u6211\u5011\u4e5f\u5617\u8a66\u4f7f\u7528\u7d44\u5408\u5f0f\u7684\u76e3\u7763\u5f0f\u6a5f\u5668\u5b78\u7fd2\u6f14\u7b97\u6cd5(ensemble approach)\uff0c\uf92d\uf901\u9032 \u4e00\u6b65\u5f97\u5230\uf901\u9ad8\u7684\u6548\u80fd\uff0c\u4ee5\u4e0b\u6211\u5011\u5206\u5225\u8a73\u7d30\u4ecb\u7d39\u3002 \uf9e1\u653f\u5112 \u7b49 3.1 \u57fa\u790e\u7fa9\u539f\u7279\u5fb5 \u5f9e E-HowNet \u62bd\u53d6\u7684\u7279\u5fb5\u7a31\u4e4b\u70ba\u57fa\u790e\u7fa9\u539f\u7279\u5fb5\uff0c\u4e5f\u5c31\u662f\u5c0d\u6bcf\u4e00\u500b E-HowNet \u7684\u8a5e\u5f59 i\uff0c\u7528 \u4e00\u5411\uf97e V i = (w i,j ) = (w i,1 , w i,2 , \u2026, w i,n ) \u8868\u793a\uff0c\u5176\u4e2d n \u70ba\u5411\uf97e\u7684\u7dad\ufa01\u3002 \u7531\u65bc\u6bcf\u4e00\u8a5e\u5f59\u7684\u6bcf\u4e00\u500b\u8a9e\u610f(sense)\u90fd\u6709\u4e00\u500b\u7d50\u69cb\u5316\u7684\u5b9a\u7fa9\u5f0f\uff0c\u800c\u4e14\u5b9a\u7fa9\u5f0f\u4e2d\u90fd\u7528 \u7fa9\u539f\uf92d\u9032\ufa08\u5b9a\u7fa9\uff0c\u516c\u5f0f (1) \u5b9a\u7fa9 V i \u4e2d\u6bcf\u500b\u7279\u5fb5\u7684\u6b0a\u91cd\uff1a \u23a9 \u23a8 \u23a7 = j \u7fa9\u539f , 0 j \u7fa9\u539f \u4e2d i \u5b9a\u7fa9\u5f0f , 1 , \uf967\u51fa\u73fe \u51fa\u73fe \u5982\u679c j i w (1) \u4ee5\u5716 1\u300c\u6c7d\u6cb9\u300d\u9019\u500b\u8a5e\u5f59\u70ba\uf9b5\uff0c\u5176\u5b9a\u7fa9\u5f0f\u4e2d\u51fa\u73fe\uf9ba\u7fa9\u539f material\uff0c\u6240\u4ee5\u5b83\u7684\u503c w \u6c7d\u6cb9 , material \u5c31\u6703\u662f 1\uff0c\u5176\u4ed6\u6c92\u51fa\u73fe\u7684\u7fa9\u539f\uff0c\u503c\u5c31\u6703\u662f 0\u3002\u6211\u5011\u5171\u4f7f\u7528\uf9ba 2567 \u500b\u7fa9\u539f\uf92d\u7576\u7279\u5fb5\u3002 \u5ee3\u7fa9\u77e5\u7db2\u7684\u8a5e\u5f59\u6709\u6b67\uf962\u6027\uff0c\u4e5f\u5c31\u662f\u6bcf\u500b\u8a5e\u5f59\u53ef\u80fd\u6709\u8a31\u591a\u8a9e\u610f\u3002\u800c\u8a5e\u5f59\u7684\u7b2c\u4e00\u500b\u8a9e\u610f\uff0c \u662f\u51fa\u73fe\u983b\uf961\u6700\u9ad8\u7684\u8a9e\u610f(\u9664\uf9ba\u56db\u500b\u8a5e\u5f59\uf9b5\u5916)\uff0c\u6240\u4ee5\u6211\u5011\u7528\u8a5e\u5f59\u7684\u7b2c\u4e00\u500b\u8a9e\u610f\uf92d\u62bd\u53d6\u7279 \u5fb5\u3002\u53ea\u5f9e\u8a5e\u5f59\u7684\u4e00\u500b\u8a9e\u610f\u62bd\u53d6\u7279\u5fb5\uff0c\u800c\uf967\u628a\u8a72\u8a5e\u5f59\u6240\u6709\u7684\u8a9e\u610f\u653e\u5728\u4e00\u8d77\uff0c\u4ee3\u8868\u9019\u7a2e\u65b9\u6cd5 \u53ef\u70ba\uf967\u540c\u7684\u8a9e\u610f\u7d66\u51fa\uf967\u540c\u7684\u6975\u6027\u9810\u6e2c\u3002\u53ea\u662f\u7531\u65bc\u76ee\u524d NTUSD \u6975\u6027\u6a19\u8a18\u53ea\u5230\u8a5e\u5f59\u7684\u5c64\u7d1a\uff0c \u6240\u4ee5\u7121\u6cd5\u5c0d\u8a9e\u610f\u7684\u5c64\u7d1a\u9032\ufa08\u6975\u6027\u9810\u6e2c\u3002\u4f46\u53ea\u8981\u6709\u8a9e\u610f\u5c64\u7d1a\u7684\u6975\u6027\u6a19\u8a18\uff0c\u6211\u5011\u9019\u7a2e\u505a\u6cd5\u53ef \u99ac\u4e0a\u5957\u7528\u3002 3.1.1 \u57fa\u790e\u7fa9\u539f\u7279\u5fb5\u52a0\u6b0a\u503c \u9664\uf9ba\u516c\u5f0f (1) \u7684\u65b9\u5f0f\u5916\uff0c\u6211\u5011\u53ef\u4ee5\uf9dd\u7528\uf901\u591a E-HowNet \u7684\u7279\u6027\uff0c\uf92d\u62bd\u53d6\u51fa\u6709\u7528\u7684\u8cc7\u8a0a\u3002 \u4e00\u500b\u53ef\u80fd\u7684\u65b9\u5f0f\u662f\u5b9a\u7fa9\u5f0f\u4e2d\u7684\u7d50\u69cb\uff0c\u5982\u679c\u628a\u5b9a\u7fa9\u5f0f\u5c55\u958b\uff0c\u6703\u5f97\u5230\u5982\u5716 2 \u7684\u6a39\uf9fa\u7d50\u69cb\u3002\u5728 \u9019\u6a39\uf9fa\u7d50\u69cb\u4e2d\uff0c\u7fa9\u539f\u6240\u5728\u7684\u6df1\ufa01\u662f\u4e00\u500b\u6709\u7528\u7684\u8cc7\u8a0a\uff0c\u56e0\u6b64\u6211\u5011\u4eff\u7167\uf9c7\u7fa4&\uf9e1\u7d20\u5efa(\u5218 & \uf9e1, 2002)\u7684\u516c\u5f0f\uff0c\u5c07\u6df1\ufa01\u7684\u8cc7\u8a0a\u7576\u4f5c\u6b0a\u91cd\u5f15\u5165\u516c\u5f0f (1)\uff0c\u5f97\u5230\u516c\u5f0f (2)\u3002 \u5716 2.\u300c\u5929\uf9d4\u4e4b\uf914\u300d\u5b9a\u7fa9\u5f0f\u7684\u6a39\uf9fa\u8868\u793a \u5ee3\u7fa9\u77e5\u7db2\u8a5e\u5f59\u610f\ufa0a\u6975\u6027\u7684\u9810\u6e2c 25 \u23aa \u23a9 \u23aa \u23a8 \u23a7 \u00d7 + = j \u7fa9\u539f , 0 j \u7fa9\u539f \u4e2d i \u5b9a\u7fa9\u5f0f , 1 1 , , \uf967\u51fa\u73fe \u51fa\u73fe \u5982\u679c j i j i d w \u03b1 (2) \u516c\u5f0f (2) \u4e2d\uff0c\u03b1 \u662f\u53ef\u8abf\u7684\uf96b\uf969\uff0c j i d , \u662f\u8a5e\u5f59 i \u8ddf\u7fa9\u539f j \u7684\u8ddd\uf9ea\uff0c\u9019\u53ef\u7528\u7fa9\u539f j \u7684\u6df1\ufa01\u8868 \u793a\u3002\u8abf\u6574\u516c\u5f0f (2) \u4e2d\u7684 \u03b1 \uff0c\u8b93\u6211\u5011\u53ef\u4ee5\u5be6\u9a57\u90a3\u4e00\u7a2e\u65b9\u5f0f\uff0c\u624d\u61c9\u7d66\u8f03\u9ad8\u7684\u6b0a\u91cd\uff1a (\u53ef\u80fd\u4e00) \u03b1 < 0 : \u6df1\ufa01\u8d8a\u6df1\uff0c\u8868\u793a\u8a72\u7fa9\u539f\u6709\u8f03\u591a\u8cc7\u8a0a\uff0c\u61c9\u7d66\u8f03\u9ad8\u6b0a\u91cd\u3002 (\u53ef\u80fd\u4e8c) \u03b1 > 0 : \u6df1\ufa01\u8d8a\u6df1\uff0c\u8868\u793a\u8a72\u7fa9\u539f\u6709\u8f03\u5c11\u8cc7\u8a0a\uff0c\u61c9\u7d66\u8f03\u5c11\u6b0a\u91cd\u3002 \u7531\u65bc \u03b1 < 0 \u6642\uff0cw i,j \u53ef\u80fd\u8b8a\u70ba\u8ca0\u503c\uff0c\u6240\u4ee5\u6700\u5c0f\u7684 \u03b1 \u8a2d\u70ba \u22120.05\u3002\u53e6\u5916\uff0c\u7576 \u03b1 = 0\uff0c\u516c\u5f0f (2) \u6703\u7b49\u65bc\u516c\u5f0f (1)\uff0c\u6240\u4ee5\u6211\u5011\u5728\u505a\u5be6\u9a57\u6642\uff0c\u53ea\u8981\u4f7f\u7528\u516c\u5f0f (2) \u5373\u53ef\u3002 3.1.2 \u52a0\u5165\u5426\u5b9a\u95dc\u4fc2\u8abf\u6574\u7279\u5fb5\u7684\u52a0\u6b0a\u503c \u5728\u8a08\u7b97\u7fa9\u539f\u6df1\ufa01\u6642\uff0c\u53ef\u80fd\u6703\u7d93\u904e\u5e36\u6709\u5426\u5b9a\u610f\u7fa9\u7684\u95dc\u4fc2\uff0c\uf9b5\u5982\u300c\u4e00\u4e8b\u7121\u6210\u300d\u5b9a\u7fa9\u5f0f\u4e2d\u6709 \u300c{not({succeed|\u6210\u529f})}\u300d\uff0c\u53ef\u4ee5\u767c\u73fe succeed \u88ab not \u6240\u4fee\u98fe\u3002\u9019\u6642\uff0c\u7fa9\u539f succeed \u7684\u6b0a\u91cd \u7528\u8ca0\u503c\uf92d\u8868\u793a\u53ef\u80fd\u6703\uf901\u597d\uff0c\u56e0\u6b64\u6211\u5011\u5c07\u5426\u5b9a\u7684\u6982\uf9a3\u5f15\u5165\u516c\u5f0f (3) \u5982\u4e0b\uff1a \u23aa \u23a9 \u23aa \u23a8 \u23a7 \u00d7 + = j \u7fa9\u539f , 0 j \u7fa9\u539f \u4e2d i \u5b9a\u7fa9\u5f0f , 1 , , , \uf967\u51fa\u73fe \u51fa\u73fe \u5982\u679c j i j i j i d Neg w \u03b1 (3) \u5176\u4e2d\uff0c j i Neg , \u8868\u793a\u7fa9\u539f j \u662f\u5426\u6709\u88ab\u5426\u5b9a\u610f\u7fa9\u7684\u95dc\u4fc2\u6240\u4fee\u98fe\uff0c\uf974\u6709\u5247 j i Neg , \u70ba \u2212 1\uff0c \uf974\u7121\u5247 j i Neg , \u70ba",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "\u7dd2\uf941",
"sec_num": "1."
},
{
"text": "V i = (c i,j ) = (c i,1 , c i,2 ,\u2026, c i,m )\u3002\u5176\u4e2d\uff0cm \u662f\u7279\u5fb5\u96c6\u5408\u7684 \u5927\u5c0f\uff0cc i,j \u662f\u300c\u8a5e\u5f59",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "\u7dd2\uf941",
"sec_num": "1."
},
{
"text": "\u7528 V i = (c i,1 , c i,2 ,\u2026, c i,m ) \u7684\u65b9\u5f0f\uf92d\u8868\u793a\u7684\u7f3a\u9ede\uff0c\u662f c i,j \u7684\u503c\u8b8a\u5316\u7684\u7bc4\u570d\u6703\u975e\u5e38\u5927\uff0c\u6700\u5c0f\u70ba 40\uff0c\u6700\u5927\u6703\u5230\u4e0a\u5343\u842c\u3002\u9019\u5728\u6a5f\u5668\u5b78\u7fd2\u4e2d\uff0c\u901a\u5e38\u9700\u8981\u505a\u9032\u4e00\u6b65\u7684\u8655\uf9e4\u624d\u6703\u6709\u6bd4\u8f03\u597d\u7684\u7d50\u679c\u3002 \u6211 \u5011 \u5be6 \u9a57 \uf9ba \uf978 \u500b \uf967 \u540c \u7684 \u65b9 \u6cd5 \uf92d \u8655 \uf9e4 \u9019 \u4e00 \u554f \u984c \uff1a \u7b2c \u4e00 \u7a2e \u662f \u4e00 \u822c \u7684 \u9918 \u5f26 \u6a19 \u6e96 \u5316 (cosine-normalization)\uff0c\u5c07\u539f\u672c\u7684\u5411\uf97e V i \u7528\u516c\u5f0f (4) \u8655\uf9e4\uff1b\u7b2c\u4e8c\u7a2e\u662f Esuli & Sebastiani (2005) \u6240\u63d0\u7684\u9918\u5f26\u6a19\u6e96\u5316 TFIDF (cosine-normalized TF-IDF)\uff0c\u4ed6\u5011\u7528\u8a72\u65b9\u6cd5\uf92d\u8655\uf9e4 WordNet \u4e2d\u7684\u8a5e\u5f59\u7684\u6b0a\u91cd\uff0c\u5982\u516c\u5f0f (5) \u6240\u8ff0\u3002 m m k k i i i c V V CosNorm \u211c \u2208 = \u2211 \u2264 \u2264 1 2 , ) ( (4) m m k k i i i tfidf TFIDF TFIDF CosNorm \u211c \u2208 = \u2211 \u2264 \u2264 1 2 , ) ( ) ,..., , ( , 2 , 1 , m i i i i tfidf tfidf tfidf TFIDF = j j i j i idf tf tfidf * , , = \u2211 \u2208 = = D k j k j i j i j i c c j c tf , , , , \u7e3d\u51fa\u73fe\u6b21\uf969 \u7279\u5fb5 } , 0 : { log ) log( , 1 D i c i D df idf j i j j \u2208 \u2200 > = = \u2212 (5) \u516c\u5f0f (5)\u4e2d D \u8868\u793a\u6587\u4ef6\u7684\u96c6\u5408\uff0c\u6b64\u8655\u628a\u8a5e\u5f59 i \u7576\u6210\u6587\u4ef6\uff0c\u7279\u5fb5 j \u7576\u6210 term\u3002 \u516c\u5f0f (4) \u7684\u6a19\u6e96\u5316\u53ef\u4ee5\u8b93\u6240\u6709\u8a5e\u5f59\u7684\u5411\uf97e\u7b49\u9577\uff0c\u6d88\u6389\u6b21\uf969\u8b8a\u5316\u904e\u5927\u7684\u7f3a\u9ede\u3002\u516c\u5f0f (5) \u7684\u60f3\u6cd5\u5247\u8a8d\u70ba\u7279\u5fb5 j \u7684\u6b0a\u91cd\uff0c\u61c9\u8a72\u5148\u8de8\u8a5e\u5f59\u9032\ufa08\u6a19\u6e96\u5316(normalization)\uff0c\u6240\u4ee5 tf i , j \u6703 \u9664\u4ee5\u7279\u5fb5 j \u7684\u7e3d\u51fa\u73fe\u6b21\uf969\uff0c\u53e6\u5916\u518d\u8003\u616e\u7279\u5fb5 j \u7684\u7a00\u6709\ufa01\uff0c\u6240\u4ee5\u4e58\u4e0a idf j \uff0c\u6700\u5f8c\u518d\u8b93\u6240 \u6709\u8a5e\u5f59\u7684\u5411\uf97e\u7b49\u9577\u3002\u6211\u5011\u6703\u5728\u5f8c\u9762\u7684\u5be6\u9a57\u4e2d\uff0c\u6bd4\u8f03\u9019\uf978\u7a2e\uf967\u540c\u6b0a\u91cd\u8655\uf9e4\u65b9\u5f0f\u7684\u6548\u80fd\u3002 3.3 \uf967\u540c\u7279\u5fb5\u7684\u7d44\u5408 \u6211\u5011\u7528\uf9ba\u57fa\u790e\u7fa9\u539f\u7279\u5fb5 (w i,1 , w i,2 ,\u2026, w i,n ) = (w i,j ) \uff0c\u53ca\u8a9e\u7bc7\u7279\u5fb5 (c i,1 , c i,2 ,\u2026, c i,m ) = (c i,j ) \uf92d \u8868\u793a\u8a5e\u5f59 i\u3002\u5982\u679c\u60f3\u540c\u6642\u4f7f\u7528\u9019\uf978\u7a2e\u7279\u5fb5\u4e2d\u7684\u8cc7\u8a0a\uff0c\u4e00\u7a2e\u76f4\u89c0\u7684\u65b9\u5f0f\uff0c\u662f\u5c07\uf978\u7a2e\u7279\u5fb5\u8868 \u793a\u65b9\u5f0f\u6df7\u5408\uff0c\u7528 V i = (w i,1 , w i,2 ,\u2026, w i,n ,c i,1 , c i,2 ,\u2026, c i,m ) \uf92d\u8868\u793a\u3002\u7531\u65bc\u57fa\u790e\u7fa9\u539f\u7279\u5fb5\u53ca\u8a9e\u7bc7 \u7279\u5fb5\u90fd\u6709\u8a31\u591a\uf967\u540c\u7684\u8b8a\u5f62\uff0c\u6211\u5011\u7121\u6cd5\u4e00\u4e00\u5617\u8a66\u6240\u6709\u53ef\u80fd\u7684\u7d44\u5408\uff0c\u6240\u4ee5\u6703\u5148\u5206\u5225\u7528\u5be6\u9a57\u627e \u51fa\u6700\u597d\u7684\u57fa\u790e\u7fa9\u539f\u7279\u5fb5 (w i,j ) \u53ca\u8a9e\u7bc7\u7279\u5fb5 (c i,j )",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "\u7dd2\uf941",
"sec_num": "1."
},
{
"text": "\uf92d\u6e2c\u8a66\u5206\uf9d0\u5668\u7684\u6548\u80fd\u5dee\u8ddd\u662f\u5426\u70ba\u986f\u8457\uff0c\u986f\u8457\u6c34 \u6e96\u8a2d\u5b9a\u70ba 0.95\u3002 McNemar \u6aa2\u5b9a\u5c07\u6e2c\u8a66\u8cc7\uf9be\u4f9d\u7167\uf978\u500b\u5206\uf9d0\u5668 (\u4ee5\u4e0b\u7a31\u70ba\u5206\uf9d0\u5668 A \u8207\u5206\uf9d0\u5668 B) \u7684\u6a19\u8a18\uff0c \u5206\u6210\u56db\u7d44\u4e26\u8a08\uf969\u3002\u5176\u4e2d\u6e2c\u8a66\u6a23\u672c\uf969\u5373\u70ba\u4e0b\u9762 n 1,1 \u3001n 0,1 \u3001n 1,0 \u3001n 0,0 \u56db\u500b\uf969\u5b57\u7684\u7e3d\u5408\uff0c\u5728\u865b \u7121\u5047\u8a2d(null hypothesis)\u4e2d\uff0c\uf978\u500b\u5206\uf9d0\u5668\u61c9\u5177\u6709\u76f8\u540c\u7684\u932f\u8aa4\uf961\uff0c\u4e5f\u5c31\u662f n 0,1 =n 1,0 \u3002 n 1,1 \uff1a \u5206\uf9d0\u5668 A \u8207\u5206\uf9d0\u5668 B \u7686\u6b63\u78ba\u6a19\u8a18 \u7684\u6a23\u672c\uf969 n 0,1 \uff1a \u5206\uf9d0\u5668 A \u6a19\u8a18\u932f\u8aa4\uff0c\u4f46\u5206\uf9d0\u5668 B \u6a19\u8a18\u6b63\u78ba\u7684\u6a23\u672c\uf969 n 1,0 \uff1a \u5206\uf9d0\u5668 B \u6a19\u8a18\u932f\u8aa4\uff0c\u4f46\u5206\uf9d0\u5668 A \u6a19\u8a18\u6b63\u78ba\u7684\u6a23\u672c\uf969 n 0,0 \uff1a \u5206\uf9d0\u5668 A \u8207\u5206\uf9d0\u5668 B \u7686\u932f\u8aa4\u6a19\u8a18",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "\u7dd2\uf941",
"sec_num": "1."
},
{
"text": "http://ehownet.iis.sinica.edu.tw/",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
}
],
"back_matter": [],
"bib_entries": {
"BIBREF0": {
"ref_id": "b0",
"title": "Creating robust supervised classifiers via web-scale N-gram data",
"authors": [
{
"first": "S",
"middle": [],
"last": "Bergsma",
"suffix": ""
},
{
"first": "E",
"middle": [],
"last": "Pitler",
"suffix": ""
},
{
"first": "D",
"middle": [],
"last": "Lin",
"suffix": ""
}
],
"year": 2010,
"venue": "Proceedings of the 48th Annual Meeting of the Association for Computational Linguistics",
"volume": "",
"issue": "",
"pages": "865--874",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Bergsma, S., Pitler, E., & Lin, D. (2010). Creating robust supervised classifiers via web-scale N-gram data. In Proceedings of the 48th Annual Meeting of the Association for Computational Linguistics, 865-874.",
"links": null
},
"BIBREF1": {
"ref_id": "b1",
"title": "LIBSVM: a library for support vector machines",
"authors": [
{
"first": "C.-C",
"middle": [],
"last": "Chang",
"suffix": ""
},
{
"first": "C.-J",
"middle": [],
"last": "Lin",
"suffix": ""
}
],
"year": 2001,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Chang, C.-C., & Lin, C.-J. (2001). LIBSVM: a library for support vector machines.",
"links": null
},
"BIBREF2": {
"ref_id": "b2",
"title": "Approximate statistical tests for comparing supervised classification learning algorithms",
"authors": [
{
"first": "T",
"middle": [
"G"
],
"last": "Dietterich",
"suffix": ""
}
],
"year": 1998,
"venue": "Neural computation",
"volume": "10",
"issue": "",
"pages": "1895--1923",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Dietterich, T. G. (1998). Approximate statistical tests for comparing supervised classification learning algorithms. In Neural computation, 10(7), 1895-1923.",
"links": null
},
"BIBREF3": {
"ref_id": "b3",
"title": "HowNet and the Computation of Meaning",
"authors": [
{
"first": "Z",
"middle": [],
"last": "Dong",
"suffix": ""
},
{
"first": "Q",
"middle": [],
"last": "Dong",
"suffix": ""
}
],
"year": 2006,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Dong, Z., & Dong, Q. (2006). HowNet and the Computation of Meaning. World Scientific.",
"links": null
},
"BIBREF4": {
"ref_id": "b4",
"title": "Determining the semantic orientation of terms through gloss classification",
"authors": [
{
"first": "A",
"middle": [],
"last": "Esuli",
"suffix": ""
},
{
"first": "F",
"middle": [],
"last": "Sebastiani",
"suffix": ""
}
],
"year": 2005,
"venue": "Proceedings of CIKM-05",
"volume": "",
"issue": "",
"pages": "617--624",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Esuli, A., & Sebastiani, F. (2005). Determining the semantic orientation of terms through gloss classification, In Proceedings of CIKM-05, 617-624.",
"links": null
},
"BIBREF5": {
"ref_id": "b5",
"title": "SentiWordNet: A publicly available lexical resource for opinion mining",
"authors": [
{
"first": "A",
"middle": [],
"last": "Esuli",
"suffix": ""
},
{
"first": "F",
"middle": [],
"last": "Sebastiani",
"suffix": ""
}
],
"year": 2006,
"venue": "Proceedings of the 5th Conference on Language Resources and Evaluation (LREC 06",
"volume": "",
"issue": "",
"pages": "417--422",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Esuli, A., & Sebastiani, F. (2006a). SentiWordNet: A publicly available lexical resource for opinion mining. In Proceedings of the 5th Conference on Language Resources and Evaluation (LREC 06) , 417-422.",
"links": null
},
"BIBREF6": {
"ref_id": "b6",
"title": "Determining term subjectivity and term orientation for opinion mining",
"authors": [
{
"first": "A",
"middle": [],
"last": "Esuli",
"suffix": ""
},
{
"first": "F",
"middle": [],
"last": "Sebastiani",
"suffix": ""
}
],
"year": 2006,
"venue": "Proceedings of the 11th Conference of the European Chapter of the Association for Computational Linguistics",
"volume": "",
"issue": "",
"pages": "193--200",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Esuli, A., & Sebastiani, F. (2006b). Determining term subjectivity and term orientation for opinion mining. In Proceedings of the 11th Conference of the European Chapter of the Association for Computational Linguistics, 193-200.",
"links": null
},
"BIBREF7": {
"ref_id": "b7",
"title": "Efficiently identifying semantic orientation algorithm for Chinese words",
"authors": [
{
"first": "Z",
"middle": [],
"last": "Han",
"suffix": ""
},
{
"first": "Q",
"middle": [],
"last": "Mo",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Zuo",
"suffix": ""
},
{
"first": "D",
"middle": [],
"last": "Duan",
"suffix": ""
}
],
"year": 2010,
"venue": "International Conference on Computer Application and System Modeling",
"volume": "2",
"issue": "",
"pages": "260--264",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Han, Z., Mo, Q., Zuo, M., & Duan, D. (2010). Efficiently identifying semantic orientation algorithm for Chinese words. In International Conference on Computer Application and System Modeling, Vol. 2, 260-264.",
"links": null
},
"BIBREF8": {
"ref_id": "b8",
"title": "Predicting the semantic orientation of adjectives",
"authors": [
{
"first": "V",
"middle": [],
"last": "Hatzivassiloglou",
"suffix": ""
},
{
"first": "K",
"middle": [
"R"
],
"last": "Mckeown",
"suffix": ""
}
],
"year": 1997,
"venue": "Proceedings of ACL-97",
"volume": "",
"issue": "",
"pages": "174--181",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Hatzivassiloglou, V., & McKeown, K. R. (1997). Predicting the semantic orientation of adjectives. In Proceedings of ACL-97, 174-181.",
"links": null
},
"BIBREF10": {
"ref_id": "b10",
"title": "Using WordNet to measure semantic orientation of adjectives",
"authors": [
{
"first": "J",
"middle": [],
"last": "Kamps",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Marx",
"suffix": ""
},
{
"first": "R",
"middle": [
"J"
],
"last": "Mokken",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "De Rijke",
"suffix": ""
}
],
"year": 2004,
"venue": "Proceedings of the 4th International Conference on Language Resources and Evaluation (LREC'04)",
"volume": "IV",
"issue": "",
"pages": "1115--1118",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Kamps, J., Marx, M., Mokken, R. J., & De Rijke, M. (2004). Using WordNet to measure semantic orientation of adjectives. In Proceedings of the 4th International Conference on Language Resources and Evaluation (LREC'04), Vol. IV, 1115-1118.",
"links": null
},
"BIBREF11": {
"ref_id": "b11",
"title": "Mining opinions from the Web: Beyond relevance retrieval",
"authors": [
{
"first": "L.-W",
"middle": [],
"last": "Ku",
"suffix": ""
},
{
"first": "H.-H",
"middle": [],
"last": "Chen",
"suffix": ""
}
],
"year": 2007,
"venue": "In Journal of the American Society for Information Science and Technology",
"volume": "58",
"issue": "12",
"pages": "1838--1850",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Ku, L.-W., & Chen, H.-H. (2007). Mining opinions from the Web: Beyond relevance retrieval. In Journal of the American Society for Information Science and Technology, 58(12), 1838-1850.",
"links": null
},
"BIBREF12": {
"ref_id": "b12",
"title": "Words semantic orientation classification based on HowNet",
"authors": [
{
"first": "D",
"middle": [],
"last": "Li",
"suffix": ""
},
{
"first": "Y.-Tao",
"middle": [],
"last": "Ma",
"suffix": ""
},
{
"first": "J.-Li",
"middle": [],
"last": "Guo",
"suffix": ""
}
],
"year": 2009,
"venue": "The Journal of China Universities of Posts and Telecommunications",
"volume": "16",
"issue": "",
"pages": "106--110",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Li, D., Ma, Y.-tao, & Guo, J.-li. (2009). Words semantic orientation classification based on HowNet. In The Journal of China Universities of Posts and Telecommunications, 16(1), 106-110.",
"links": null
},
"BIBREF13": {
"ref_id": "b13",
"title": "Chinese Web 5-gram Version 1. Linguistic Data Consortium",
"authors": [
{
"first": "F",
"middle": [],
"last": "Liu",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Yang",
"suffix": ""
},
{
"first": "D",
"middle": [],
"last": "Lin",
"suffix": ""
}
],
"year": 2010,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Liu, F., Yang, M., & Lin, D. (2010). Chinese Web 5-gram Version 1. Linguistic Data Consortium, Philadelphia.",
"links": null
},
"BIBREF14": {
"ref_id": "b14",
"title": "Learning Chinese polarity lexicons by integration of graph models and morphological features",
"authors": [
{
"first": "B",
"middle": [],
"last": "Lu",
"suffix": ""
},
{
"first": "Y",
"middle": [],
"last": "Song",
"suffix": ""
},
{
"first": "X",
"middle": [],
"last": "Zhang",
"suffix": ""
},
{
"first": "B",
"middle": [],
"last": "Tsou",
"suffix": ""
}
],
"year": 2010,
"venue": "Information Retrieval Technology: 6th Asia Information Retrieval Societies Conference",
"volume": "2010",
"issue": "",
"pages": "466--477",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Lu, B., Song, Y., Zhang, X., & Tsou, B. (2010). Learning Chinese polarity lexicons by integration of graph models and morphological features. In Information Retrieval Technology: 6th Asia Information Retrieval Societies Conference, AIRS 2010, 466-477.",
"links": null
},
"BIBREF15": {
"ref_id": "b15",
"title": "Measuring praise and criticism: Inference of semantic orientation from association",
"authors": [
{
"first": "P",
"middle": [
"D"
],
"last": "Turney",
"suffix": ""
},
{
"first": "M",
"middle": [
"L"
],
"last": "Littman",
"suffix": ""
}
],
"year": 2003,
"venue": "In ACM Transactions on Information Systems",
"volume": "21",
"issue": "",
"pages": "315--346",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Turney, P. D., & Littman, M. L. (2003). Measuring praise and criticism: Inference of semantic orientation from association. In ACM Transactions on Information Systems, 21, 315-346.",
"links": null
},
"BIBREF16": {
"ref_id": "b16",
"title": "Using bilingual lexicon to judge sentiment orientation of Chinese words",
"authors": [
{
"first": "J",
"middle": [],
"last": "Yao",
"suffix": ""
},
{
"first": "G",
"middle": [],
"last": "Wu",
"suffix": ""
},
{
"first": "J",
"middle": [],
"last": "Liu",
"suffix": ""
},
{
"first": "Y",
"middle": [],
"last": "Zheng",
"suffix": ""
}
],
"year": 2006,
"venue": "Proceedings of the Sixth IEEE International Conference on Computer and Information Technology",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Yao, J., Wu, G., Liu, J., & Zheng, Y. (2006). Using bilingual lexicon to judge sentiment orientation of Chinese words, In Proceedings of the Sixth IEEE International Conference on Computer and Information Technology, p. 38.",
"links": null
},
"BIBREF18": {
"ref_id": "b18",
"title": "Morpheme-based derivation of bipolar semantic orientation of Chinese words",
"authors": [],
"year": null,
"venue": "Proceedings of the 20th international conference on Computational Linguistics",
"volume": "",
"issue": "",
"pages": "1008--1014",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Morpheme-based derivation of bipolar semantic orientation of Chinese words, In Proceedings of the 20th international conference on Computational Linguistics, 1008-1014.",
"links": null
}
},
"ref_entries": {
"FIGREF0": {
"type_str": "figure",
"text": "",
"num": null,
"uris": null
},
"TABREF0": {
"content": "<table><tr><td>(Extended-HowNet Ontology)\uff0c\u4e26\u7528\u9019\u4e9b\u65b0\u7684\u8a9e\u7fa9\u7fa9\u539f\uff0c\u4ee5\u7d50\u69cb\u5316\u7684\u5f62\u5f0f\uf92d\u5b9a\u7fa9\u8a5e\u689d\uff0c</td></tr><tr><td>\u8a5e\u689d\u5b9a\u7fa9\u5f0f\u7684\uf9b5\u5b50\u5982\u5716 1\u3002</td></tr><tr><td>\u6709\u95dc\u60c5\u7dd2\u5c6c\u6027\u6a19\u8a18\u7684\u7814\u7a76\uff0c\u6211\u5011\u5206\u70ba\u82f1\u6587\u53ca\u4e2d\u6587\uf92d\u8a0e\uf941\u3002\u5728\u82f1\u6587\u65b9\u9762\uff0c\u6700\u65e9\u662f\u7531</td></tr><tr><td>Hatzivassiloglou & McKeown(1997) \u5728 1997 \uf98e\u91dd\u5c0d\u5f62\u5bb9\u8a5e\u6240\u505a\u7684\u7814\u7a76\uff0c\u4ed6\u5011\u6240\u7528\u7684\u5f62\u5bb9</td></tr><tr><td>\u8a5e\u5206\u5225\u6709\u6b63\u9762\u8a5e 657 \u500b\u53ca\u8ca0\u9762\u8a5e 679 \u500b\uff0c\u8a72\uf941\u6587\u4f9d\u64da\uf967\u540c\u7684\u5be6\u9a57\u8a2d\u5b9a\uff0c\u76e3\u7763\u5f0f\u6a5f\u5668\u5b78\u7fd2</td></tr><tr><td>\u7684\u6e96\u78ba\uf961(Accuracy)\u7531 82% \u5230 90%\u3002\u4e4b\u5f8c\uf9d3\u7e8c\u6709\uf967\u540c\u7684\u7814\u7a76\uff0c\u6240\u7528\u591a\u70ba\u534a\u76e3\u7763\u5f0f\u6a5f</td></tr><tr><td>\u5668\u5b78\u7fd2\u7684\u6f14\u7b97\u6cd5</td></tr><tr><td>\u8981\u5168\u90e8\u7528\u4eba\u5de5\u9032\ufa08\u6a19\u8a18\u4e4b\u6210\u672c\u592a\u9ad8\u3002\u56e0\u6b64\uff0c\u901a\u5e38\u7684\u4f5c\u6cd5\u662f\u5c11\uf97e\u6a19\u8a18\u4e00\u4e9b\u8a5e\u5f59\uff0c\u518d\u7528\u6a5f\u5668</td></tr><tr><td>\u5b78\u7fd2\u65b9\u6cd5\uff0c\u70ba\u5269\u4e0b\u7684\u8a5e\u5f59\u9032\ufa08\u81ea\u52d5\u6a19\u8a18\uff0c\u96d6\u7136\u81ea\u52d5\u6a19\u8a18\u7684\u6e96\u78ba\uf961\uf967\u5982\u4eba\u5de5\u6a19\u8a18\uff0c\u4f46\u5c0d\u4e00</td></tr><tr><td>\u822c\u61c9\u7528\u6709\u67d0\u7a2e\u7a0b\ufa01\u7684\u5e6b\u52a9\u3002</td></tr><tr><td>\u5728\u4e2d\u6587\u81ea\u7136\u8a9e\u8a00\u8655\uf9e4\uff0cNTUSD \u662f\u4e00\u90e8\u91cd\u8981\u7684\u610f\ufa0a\u8a5e\u8a5e\u5178\uff0c\u4f46\u6b64\u8a5e\u5178\u53ea\u5305\u62ec\u8a5e\u5f59\u53ca</td></tr><tr><td>\u6975\u6027\u7684\u8cc7\u8a0a\u3002\u53e6\u4e00\u65b9\u9762\uff0c\u8463\u632f\u6771\u5148\u751f\u548c\u9673\u514b\u5065\u6559\u6388\u6240\u5efa\uf9f7\u7684\u77e5\u7db2\u548c\u5ee3\u7fa9\u77e5\u7db2(Z. Dong &</td></tr><tr><td>Dong, 2006; \u9673\u514b\u5065, \u9ec3, \u65bd, & \u9673, 2004)\uff0c\u662f\u91cd\u8981\u7684\u8a9e\u610f\u8cc7\u6e90\u3002\u5c0d\u65bc\u6bcf\u500b\u8a5e\u5f59\uff0c\u90fd\u7528\u6709</td></tr><tr><td>\u9650\u7684\u7fa9\u539f\u7d66\u4e88\u7cbe\u78ba\u7684\u5b9a\u7fa9\uff0c\u4f46\u9019\u4e9b\u5b9a\u7fa9\u537b\u7f3a\u4e4f\u60c5\u7dd2\u7684\u8a9e\u610f\u6a19\u8a18\u3002\u56e0\u6b64\uff0c\u5982\u4f55\u81ea\u52d5\u70ba\u5ee3\u7fa9</td></tr><tr><td>\u77e5\u7db2\u52a0\u4e0a\u60c5\u7dd2\u6a19\u8a18\uff0c\u6210\u70ba\u4e00\u500b\u91cd\u8981\u7684\u8ab2\u984c\uff0c\u4e5f\u662f\u672c\u7814\u7a76\u7684\u76ee\u7684\u3002</td></tr><tr><td>\u672c\u7814\u7a76\u63d0\u51fa\u70ba\u5ee3\u7fa9\u77e5\u7db2\u52a0\u4e0a\u60c5\u7dd2\u6a19\u8a18\u7684\u65b9\u6cd5\uff0c\u9996\u5148\uf9dd\u7528 NTUSD \u8ddf\u5ee3\u7fa9\u77e5\u7db2\u8a5e\u5f59\u7684</td></tr><tr><td>\u4ea4\u96c6\u5efa\uf9f7\u6a19\u6e96\u7b54\u6848\u96c6\uff0c\u518d\u7531\u6a19\u6e96\u7b54\u6848\u96c6\u8a13\uf996\u51fa\u5206\uf9d0\u5668\uff0c\u70ba\u5176\u4ed6\u5ee3\u7fa9\u77e5\u7db2\u8a5e\u5f59\u9032\ufa08\u6a19\u8a18\u3002</td></tr><tr><td>\u5982\u4f55\u6709\u6548\u7684\u904b\u7528\u76e3\u7763\u5f0f\u6a5f\u5668\u5b78\u7fd2\u6f14\u7b97\u6cd5\uff0c\u5982\u4f55\u70ba\u8a5e\u5f59\u62bd\u53d6\u51fa\u6709\u7528\u7684\u7279\u5fb5\uff0c\u662f\u4e3b\u8981\u7684\u6311\u6230</td></tr><tr><td>\u8b70\u984c\u3002\u5728\u6b64\u7814\u7a76\u4e2d\uff0c\u6211\u5011\u6709\u7cfb\u7d71\u7684\u5617\u8a66\u62bd\u53d6\u5404\u7a2e\uf967\u540c\u7684\u8a5e\u5f59\u7279\u5fb5\uff0c\u6700\u5f8c\u5f97\u5230\u9ad8\u6e96\u78ba\uf961\u7684</td></tr><tr><td>\u4e8c\u5143\u5206\uf9d0\u5668(binary classifiers)\u7528\u4ee5\u81ea\u52d5\u6a19\u8a18\u6b63\u8ca0\u9762\u60c5\u7dd2\u6a19\u8a18\u3002</td></tr><tr><td>\u7b2c\u4e8c\u7bc0\u4ecb\u7d39\u5ee3\u7fa9\u77e5\u7db2\u3001\u53ca\u82f1\u6587\u548c\u4e2d\u6587\u76f8\u95dc\u7684\u60c5\u7dd2\u5c6c\u6027\u6a19\u8a18\u7814\u7a76\uff0c\u7b2c\u4e09\u7bc0\u4ecb\u7d39\u5f9e</td></tr><tr><td>E-HowNet \u53ca \u8ddf\u77e5\u7db2\uf99a\u7d50\uff0c\u4e26\u4f5c\uf9ba\u4e00\u4e9b\u4fee\u6539\uff0c\u6700\u5f8c\u5f62\u6210\u5ee3\u7fa9\u77e5\u7db2(Extended-HowNet, E-HowNet 1 )\u3002</td></tr><tr><td>\u8a5e\u5eab\u5c0f\u7d44\u4fee\u6539\u4e26\u64f4\u5c55\u77e5\u7db2\u539f\u5148\u7684\u8a9e\u7fa9\u7fa9\u539f\u89d2\u8272\u77e5\uf9fc\u672c\u9ad4\uff0c\u5efa\u69cb\u51fa\u5ee3\u7fa9\u77e5\u7db2\u77e5\uf9fc\u672c\u9ad4</td></tr></table>",
"num": null,
"type_str": "table",
"text": "\u3002\u4f46\u73fe\u6709\u8cc7\u6e90 \u7684\u8a9e\u5f59\uf97e\u901a\u5e38\u5f88\u5927\uff0c\u5982 WordNet 3.0 \u5c31\u5305\u62ec 206,941 \u500b\uf967\u540c\u7684\u82f1\u6587\u5b57\u7fa9 (word-sense pair) \uff0c Google Chinese Web 5-gram \u62bd\u53d6\u7279\u5fb5\u7684\u65b9\u6cd5\uff0c\u7b2c\u56db\u7bc0\u5448\u73fe\u5404\u7a2e\u5be6\u9a57\u7684\u7d50\u679c \u53ca\u5206\u6790\uff0c\u5305\u62ec\u8ddf NTUSD \u4eba\u5de5\u6a19\u8a18\u7684\u6bd4\u8f03\uff0c\u6700\u5f8c\u7e3d\u7d50\uf941\u6587\u7684\u6210\u679c\u3002 2. \u76f8\u95dc\u7814\u7a76 \u8463\u632f\u6771\u5148\u751f\u65bc 1998 \uf98e\u5275\u5efa\u77e5\u7db2(HowNet)\uff0c\u4e26\u5728 2003 \uf98e\uff0c\u8ddf\u4e2d\u592e\u7814\u7a76\u9662\u8cc7\u8a0a\u6240\u8a5e\u5eab\u5c0f \u7d44\u5728 2003 \uf98e\uff0c\u5c07\u4e2d\u7814\u9662\u8a5e\u5eab\u5c0f\u7d44\u8a5e\u5178(CKIP Chinese Lexical Knowledge Base)\u7684\u8a5e\u689d",
"html": null
},
"TABREF1": {
"content": "<table><tr><td>2004 \uf98e \uf9dd \u7528 Turney &</td></tr><tr><td>Littman(2003)</td></tr></table>",
"num": null,
"type_str": "table",
"text": "\u7684\u534a\u76e3\u7763\u5f0f\u6a5f\u5668\u5b78\u7fd2\u6f14\u7b97\u6cd5\uff0c\u5728\u6b63\u9762\u8a5e 604 \u500b\u53ca\u8ca0\u9762\u8a5e 645 \u500b\u7684\u8cc7\uf9be\u96c6\u4e0a \u505a\u5be6\u9a57\uff0c\u5f97\u5230\u6700\u9ad8\u7684\u6210\u7e3e\u662f 80.23%\u7684\u7cbe\u78ba\ufa01\u53ca 85.03%\u7684\u53ec\u56de\uf961\u3002\u4e4b\u5f8c\u5f9e 2006 \u5230 2010 \uf98e\uff0c\uf9d3\u7e8c\u7684\u7814\u7a76\u4f7f\u7528\uf967\u540c\u7684\u8cc7\uf9be\u96c6\uff0c\u7528\uf967\u540c\uf9d0\u578b\u7684\u6a5f\u5668\u5b78\u7fd2\u6f14\u7b97\u6cd5\uf92d\u8655\uf9e4\u9019\u500b\u554f\u984c",
"html": null
},
"TABREF2": {
"content": "<table><tr><td>3.2.1 Google Web 5-gram\u7279\u5fb5\u62bd\u53d6</td></tr><tr><td>\u6211\u5011\u4f7f\u7528\u7279\u5fb5\u8ddf\u8a5e\u5f59\u7684\u540c\u51fa\u73fe(co-occurrence)\u6b21\uf969\u505a\u70ba\u7279\u5fb5\u503c\uff0c\u4ee5\u5716 3 \u70ba\uf9b5\uff0c\u5982\u679c\u8a5e</td></tr><tr><td>+1\u3002\u53e6\u5916\uff0c\u5982\u679c\u6a39\uf9fa\u7d50\u69cb\u4e0a\u9762\u7684\u7fa9\u539f\u88ab\u5426\u5b9a\u610f\u7fa9\u7684\u95dc\u4fc2\u6240\u4fee\u98fe\uff0c\u9019\u5426\u5b9a \u5f59\u662f\u300c\u6050\u5413\u300d\uff0c\u4ee5\u300c\u975e\u6cd5\u300d\u7576\u7279\u5fb5\u503c\uff0c\u5247\u540c\u51fa\u73fe\u6b21\uf969\u6703\u5c07\u6240\u6709\u300c\u6050\u5413\u300d\u53ca\u300c\u975e\u6cd5\u300d\u4e00\u540c</td></tr><tr><td>\u610f\u7fa9\u6703\u50b3\u905e\u5230\u4e0b\u9762\u7684\u7fa9\u539f\u3002 \u51fa\u73fe\u7684 5-gram \u6b21\uf969\u76f8\u52a0\u3002\u5728\u4e0a\u9762\u7684\uf9b5\u5b50\u4e2d\uff0c\u300c\u6050\u5413\u300d\u53ca\u300c\u975e\u6cd5\u300d\u7684\u540c\u51fa\u73fe\u6b21\uf969\u70ba</td></tr><tr><td>574+200+4463 + 705=5942 \u6b21\u3002 3.2 \u8a9e\u7bc7(context)\u7279\u5fb5 \u53e6\u5916\uff0c\u7531\u65bc\u5ee3\u7fa9\u77e5\u7db2\u8ddf Google Web 5-gram \u7684\u65b7\u8a5e\u6a19\u6e96\u4e26\uf967\u4e00\u81f4\uff0c\u6240\u4ee5\u5728\u8655\uf9e4\u6642\u628a</td></tr><tr><td>\u5ee3\u7fa9\u77e5\u7db2\u96d6\u7136\u6709\u56b4\u8b39\u7684\u5b9a\u7fa9\u5f0f\u53ef\u7528\u4ee5\u8868\u793a\u8a5e\u5f59\uff0c\u4f46\u662f\u6709\u56db\u500b\u7f3a\u9ede\uff0c\u9020\u6210\u53ea\u7528\u7fa9\u539f\u7576\u7279\u5fb5 Google Web 5-gram \u7684\u7a7a\u767d\u53bb\u6389\uff0c\u76f4\u63a5\u627e\u51fa\u300c\u8a5e\u5f59\u300d\u8ddf\u300c\u7279\u5fb5\u300d\u9019\uf978\u5b57\uf905\u662f\u5426\u540c\u6642\u51fa\u73fe\uff0c</td></tr><tr><td>\u7121\u6cd5\u6b63\u78ba\u7372\u5f97\u8a5e\u5f59\u7684\u6975\u6027\u3002 \uf92d\u8a08\u7b97\u6b21\uf969\uff0c\u9019\u6a23\u53ef\u4ee5\u907f\u514d\u65b7\u8a5e\u6a19\u6e96\uf967\u4e00\u6240\u7522\u751f\u7684\u554f\u984c\u3002\uf9b5\u5982 \u300c\u4e00\u4e8b\u7121\u6210\u300d\u5728 Google Web</td></tr><tr><td>\u7b2c\u4e00\u500b\u7f3a\u9ede\u662f\u8a5e\u5f59\u6240\u6a19\u7684\u7fa9\u539f\uf97e\u592a\u5c11\uff0c\u56e0\u70ba\u8a5e\u5f59\u662f\u7528\u4eba\u5de5\u6a19\u793a\u7fa9\u539f\uff0c\u6240\u4ee5\u7121\u6cd5\u7d66\u4e88 5-gram \u4e2d\u88ab\u65b7\u6210\u56db\u500b\u7368\uf9f7\u7684\u8a5e\uff0c\u5c07\u7a7a\u767d\u53bb\u6389\u5c31\u53ef\u4ee5\u6b63\u78ba\u6bd4\u5c0d\u5230\u3002</td></tr><tr><td>\u5f88\u591a\u6a19\u793a\u3002\u9019\u8868\u793a\u8a5e\u5f59\u64c1\u6709\u7684\u8cc7\u8a0a\uf97e\u6709\u9650\uff0c\u6703\u9020\u6210\u5206\uf9d0\u5668\u7121\u6cd5\u6709\u6548\u5b78\u7fd2\u3002\u7b2c\u4e8c\u500b\u7f3a\u9ede\u662f \u56e0\u70ba\u9019\uf9e8\u7684\u8a5e\u5f59\u96c6\u5408\u5c31\u662f\u7b49\u5f85\u6a19\u793a\u6975\u6027\u7684\u8a5e\uff0c\u6240\u4ee5\u6211\u5011\u53ea\u8981\u6307\u5b9a\u7279\u5fb5\u7684\u96c6\u5408\u5305\u62ec\u90a3</td></tr><tr><td>\u7fa9\u539f\uf969\uf97e\u592a\u5c11\uff0c\u9019\u6703\u9020\u6210\u8a9e\u7fa9\u7684\u5283\u5206\uf967\u5920\u7cbe\u78ba\uff0c\u7121\u6cd5\u986f\u793a\u51fa\u771f\u5be6\u7684\u8a9e\u7fa9\u5dee\u5225\uff0c\uf9b5\u5982\u300c\u660e \u4e9b\u8a5e\uff0c\u5c31\u53ef\u7b97\u51fa\u8868\u793a\u8a5e\u5f59 i \u7684\u5411\uf97e</td></tr><tr><td>\u54f2\u4fdd\u8eab\u300d\u8ddf\u300c\ufa0a\u98a8\u8f49\u8235\u300d\u7684\u5b9a\u7fa9\u5f0f\u90fd\u662f\u300c{sly|\u72e1}\u300d\uff0c\u4f46\u300c\u660e\u54f2\u4fdd\u8eab\u300d\u662f\u6b63\u9762\u610f\ufa0a\uff0c\u300c\ufa0a</td></tr><tr><td>\u98a8\u8f49\u8235\u300d\u537b\u662f\u8ca0\u9762\u610f\ufa0a\u3002\u7b2c\u4e09\u500b\u7f3a\u9ede\u662f\u5ee3\u7fa9\u77e5\u7db2\u5b9a\u7fa9\u6a19\u6e96\u7684\u5dee\uf962\uff0c\uf9b5\u5982\uff0c\u5c08\u6709\u540d\u8a5e\u5728\u5ee3</td></tr><tr><td>\u7fa9\u77e5\u7db2\u4e2d\u6703\u7528\u5ba2\u89c0\u7684\u7fa9\u539f\uf92d\u5b9a\u7fa9\uff0c\u4f46\u8a72\u5c08\u6709\u540d\u8a5e\u7d93\u904e\u4f7f\u7528\uff0c\u537b\u53ef\u80fd\u6703\u5f15\u8d77\u4eba\u7684\u6b63\u53cd\u60c5\u7dd2</td></tr><tr><td>(\u5982\u300c\u83ab\u672d\u7279\u300d\u662f\u5c08\u6709\u540d\u8a5e\uff0c\u4f46\u537b\u5e38\u7528\uf92d\u7576\u6b63\u9762\u610f\ufa0a)\uff0c\u9019\u7a2e\u5dee\uf962\u6703\u5f15\u5165\u7a0b\ufa01\uf967\u7b49\u7684\u96dc</td></tr><tr><td>\u8a0a\u5230\u5206\uf9d0\u5668\u4e2d\u3002\u7b2c\u56db\u500b\u7f3a\u9ede\u662f\u5ee3\u7fa9\u77e5\u7db2\u5c1a\u672a\u5c0d\u6240\u6709\u8a5e\u5f59\u6a19\u4e0a\u5b9a\u7fa9\u5f0f\uff0c\uf9b5\u5982\u300c\u4e7e\u6de8\u4fd0\uf918\u300d</td></tr><tr><td>\u5728\u5ee3\u7fa9\u77e5\u7db2\u53ca NTUSD \u4e2d\u90fd\u51fa\u73fe\uff0c\u4f46\u5ee3\u7fa9\u77e5\u7db2\u537b\u6c92\u6709\u6a19\u4e0a\u5b9a\u7fa9\u5f0f\u3002</td></tr></table>",
"num": null,
"type_str": "table",
"text": "\u6240\u5efa\uf9f7\u7684 Google Web \uf9e1\u653f\u5112 \u7b49 5-gram Version 1\uff0c\uf92d\u62bd\u53d6\u8a9e\u7bc7\u7279\u5fb5\u3002Google Web 5-gram \u662f Google \u5f9e\u7db2\uf937\u4e2d\u6536\u96c6\u5927\uf97e\u7684 \u7c21\u9ad4\u4e2d\u6587\u7db2\u9801\uff0c\u4e26\u7d93\u904e\u8655\uf9e4\u6240\u5efa\uf9f7\u7684\u8cc7\u6e90\u3002\u4ed6\u5011\u6536\u96c6\uf9ba 882,996,532,572 \u500b\u5b57\u7b26 (token) \uff0c \u5171 102,048,435,515 \u500b\uf906\u5b50\uff0c\u7d93\u904e\u65b7\u8a5e\u5f8c\u5efa\u6210 n-gram\u3002n-gram \u7684 n \u5f9e 1 \u5230 5\uff0c\u4e26\u4e14\u53ea \u4fdd\uf9cd\u983b\uf961\u5927\u65bc 40 \u7684 n-gram\u3002Google Web 5-gram \u7684\uf9b5\u5b50\u5982\u5716 3 \u6240\u793a\u3002",
"html": null
},
"TABREF3": {
"content": "<table><tr><td>\u5ee3\u7fa9\u77e5\u7db2\u8a5e\u5f59\u610f\ufa0a\u6975\u6027\u7684\u9810\u6e2c</td><td>27</td></tr><tr><td>3.2</td><td/></tr><tr><td>NTUSD \u5b8c\u6574\u7248\u3002</td><td/></tr></table>",
"num": null,
"type_str": "table",
"text": "i\u300d\u8ddf\u300c\u7279\u5fb5 j\u300d\u9019\uf978\u5b57\uf905\u540c\u51fa\u73fe\u7684\u6b21\uf969\u3002\u5728\u6211\u5011\u7684\u5be6\u9a57\u4e2d\uff0c\u5171\u5617\u8a66\uf9ba \u5341\u7a2e\uf967\u540c\u7684\u7279\u5fb5\u96c6\u5408\uff0c\u5206\u5225\u662f\u5ee3\u7fa9\u77e5\u7db2\u7684\u540d\u8a5e\u3001\u5ee3\u7fa9\u77e5\u7db2\u7684\u52d5\u8a5e\u3001\u5ee3\u7fa9\u77e5\u7db2\u7684\u526f\u8a5e\u3001\u5ee3 \u7fa9\u77e5\u7db2\u7684\u5f62\u5bb9\u8a5e\u3001\u5ee3\u7fa9\u77e5\u7db2\u6240\u6709\u8a5e\u5f59\u3001Google Web 5-gram \u6700\u5e38\u51fa\u73fe\u7684 5000 \u8a5e\u3001Google Web 5-gram \u6700\u5e38\u51fa\u73fe\u7684 5000 \u8a5e(\u4f46\u8a5e\u5f59\u9577\ufa01\u6700\u5c11\u70ba 2)\u3001Google Web 5-gram \u6700\u5e38\u51fa\u73fe \u7684 10000 \u8a5e\u3001Google Web 5-gram \u6700\u5e38\u51fa\u73fe\u7684 10000 \u8a5e(\u4f46\u8a5e\u5f59\u9577\ufa01\u6700\u5c11\u70ba 2)\u3001\u4ee5\u53ca",
"html": null
},
"TABREF5": {
"content": "<table><tr><td>\u7684\u6a23\u672c\uf969 McNemar \u6aa2\u5b9a\u5efa\u69cb\u5728\u5361\u65b9\u9069\u5408\ufa01\u6aa2\u5b9a(\u03c72 test goodness of fit)\u4e0a\uff0c\u6574\uf9e4\u800c\u5f97\u7684\u6aa2\u5b9a 0 , 1 1 , 0 2 0 , 1 1 , 0 ) 1 ( n n n n + \u2212 \u2212 \uff0c\u6b64\u6aa2\u5b9a\u503c\u5728 n 0,1 +n 1,0 \u5920\u5927\u7684\u6642\u5019\u6703\u8da8\u8fd1\u65bc\u81ea\u7531\ufa01\u70ba 1 \u7684\u5361\u65b9\u5206\u914d\uff0c \u56e0\u6b64\u5728\u986f\u8457\u6c34\u6e96(significant level)\u70ba 0.95 \u6642\uff0c\u6b64\u503c\uf974\u5927\u65bc \u503c\u70ba 8415 . 3 2 95 . 0 , 1 = \u03c7 \uff0c\u5247\u62d2\u7d55\u865b\u7121\u5047 \u8a2d\u3002\u6211\u5011\u7528 (McNemar \u6aa2\u5b9a\u7d50\u679c, p-value) \uf92d\u986f\u793a\u6211\u5011\u7684\u6aa2\u5b9a\u7d50\u679c\uff0c\uf9b5\u5982\u6aa2\u5b9a\u7d50\u679c (1.50, 0.22) \u8868\u793a\uff0cMcNemar \u6aa2\u5b9a\u7d50\u679c\u70ba 1.50 < 3.84\uff0c\u6240\u4ee5\u6c92\u6709\u901a\u904e McNemar \u6aa2\u5b9a\uff0cp-value \u70ba 0.22\u3002 4.2 \u57fa\u790e\u7fa9\u539f\u7279\u5fb5\u7684\u6548\u80fd 89.4% 89.6% 89.8% 90.0% Accuracy PBF PBFN \u5ee3\u7fa9\u77e5\u7db2\u8a5e\u5f59\u610f\ufa0a\u6975\u6027\u7684\u9810\u6e2c 31 \u6211\u5011\u4f7f\u7528\u4e09\u7a2e\uf967\u540c\u7684\u52a0\u6b0a\u65b9\u5f0f\u5f97\u5230\u7684\u9810\u6e2c\u6e96\u78ba\uf961\u5982\u5716 5\uff0c\u5716\u4e2d\u6211\u5011\u4e5f\u628a\u7279\u5fb5\u96c6\u7684\u7279 \u5fb5\uf969\u7531\u5de6\u81f3\u53f3\u7531\u5c0f\u5230\u5927\u6392\uf99c\u3002 \u5f9e\u5716 5 \u53ef\u4ee5\u770b\u51fa\uff0c\u6c92\u6709\u6a19\u6e96\u5316\u7684\u539f\u59cb\u983b\uf961\u7684\u6700\u4f73\u6e96\u78ba\uf961\u70ba 59.70%\uff0c\u4f7f\u7528\u7684\u7279\u5fb5\u96c6\u70ba \u300c\u5ee3\u7fa9\u77e5\u7db2\u540d\u8a5e\u300d\uff0c\u5176\u6548\u80fd\u6700\u5dee\u4e14\u5dee\u8ddd\u5f88\u5927\u3002\u9918\u5f26\u6a19\u6e96\u5316 TFIDF \u7684\u6548\u80fd\u6392\u5728\u4e2d\u9593\uff0c\u6700\u4f73 \u6e96\u78ba\uf961\u70ba 83.41%\uff0c\u4f7f\u7528\u7684\u7279\u5fb5\u96c6\u70ba\u300c\u6700\u5e38\u51fa\u73fe 10000 \u8a5e\u300d\u3002\u800c\u7d93\u904e\u9918\u5f26\u6a19\u6e96\u5316\u7684\u7279\u5fb5\u503c \u5247\u53ef\u4ee5\u5f97\u5230\u6700\u4f73\u6548\u80fd\uff0c\u5176\u6700\u4f73\u6e96\u78ba\uf961\u70ba 88.23%\uff0c\u6b64\u6642\u4f7f\u7528\u7684\u7279\u5fb5\u96c6\u70ba \u300c\u5ee3\u7fa9\u77e5\u7db2\u52d5\u8a5e\u300d \uff0c \uf9e1\u653f\u5112 \u7b49 \u63d0\u5347\uff0c\u63d0\u5347\u5f8c\u7684\u6700\u9ad8\u6e96\u78ba\uf961\u70ba 92. 3276%\uff0c\u4f7f\u7528\u300c\u5ee3\u7fa9\u77e5\u7db2\u6240\u6709\u8a5e\u5f59 All+PBFN \u03b1 = \u22120.03 \u300d \u548c\u300c\u6700\u5e38\u51fa\u73fe 10000 \u8a5e(\u9577\ufa01\u22672) F10000-2+PBFN \u03b1 = \u22120.03 \u300d\u70ba\u7279\u5fb5\u96c6\u6642\u7686\u6709\u76f8\u540c\u7684\u6e96 \u78ba\uf961\u3002\u4e0a\u5716\u4e2d\uff0c\u300c\u5ee3\u7fa9\u77e5\u7db2\u6240\u6709\u8a5e\u5f59 All\u300d\u6e96\u78ba\uf961\u5f9e 88.23%\u63d0\u5347\u81f3 92.33%\u6642\uff0c\u6b64\u5dee\u8ddd\u70ba \u986f\u8457\uff0c\u6aa2\u5b9a\u7d50\u679c (32.14, 1.4*10 -8 )\u3002 95.0% \u5ee3\u7fa9\u77e5\u7db2\u8a5e\u5f59\u610f\ufa0a\u6975\u6027\u7684\u9810\u6e2c 33 \u8868 4 \u4e2d\u7684\u7279\u5fb5\u96c6\u4ee3\u865f\u662f\u300c\u8a9e\u7bc7\u7279\u5fb5\u96c6\u4ee3\u78bc+PBFN \u03b1 = \u22120.03 \u300d\u7684\u7c21\u5beb\uff0c\u56e0\u70ba\u4f7f\u7528\u76f8\u540c\u7684 PBFN \u03b1 = \u22120.03 \uff0c\u6240\u4ee5\u5c07\u5176\u5ffd\uf976\u3002\u300c\u7e3d\u9ad4\u6548\u80fd\u300d\u662f\u6307\u5206\uf9d0\u5668\u8a13\uf996\u6642\u7684\u6574\u9ad4\u6548\u80fd\u3002\u8868\u4e2d\uff0c\u4e00\uf91d \u4e2d\u6700\u4f73\u7684\u6a19\u8a18\u6548\u80fd\u4ee5\u7c97\u9ad4\u5b57\u8868\u793a\u3002 \u8868 4 \u4e2d\u6211\u5011\u53ef\u4ee5\u767c\u73fe\uff0c\u8a13\uf996\u6642\uff0cF10000-2+PBFN \u03b1 = \u22120.03 \u6709\u6700\u9ad8\u7684\u7e3d\u9ad4\u6548\u80fd\uff0c\u5176\u5404\u8a5e \u6027\u6548\u80fd\u9664\uf9ba\u5f62\u5bb9\u8a5e\u5916\uff0c\u591a\u662f\u6700\u597d\uff1b\u8003\uf97e\u5230\u8cc7\uf9be\u96c6\u4e2d\u5f62\u5bb9\u8a5e\u7684\uf969\uf97e\u4e26\uf967\u591a\uff0c\u9019\u8868\u793a\u7d44\u5408\u591a \u500b\u5206\uf9d0\u5668\u5f8c\uff0c\u6548\u80fd\u7684\u63d0\u6607\u7a7a\u9593\u53ef\u80fd\u6709\u9650\u3002\u8868 4 \u4e2d\u53e6\u4e00\u500b\u503c\u5f97\u6ce8\u610f\u7684\u4e00\u9ede\u662f\u8a13\uf996\u8cc7\uf9be\u96c6\u7684 34 \uf9e1\u653f\u5112 \u7b49 4.6 \u76f8\u95dc\u7814\u7a76\u6548\u80fd\u6bd4\u8f03 \u6211\u5011\u7e3d\u7d50\u524d\u9762\u5404\u7a2e\uf967\u540c\u7684\u5be6\u9a57\u7d50\u679c\uff0c\u756b\u6210\u5716 7\uff0c\uf92d\u65b9\uf965\u6211\u5011\u6bd4\u8f03\u6548\u80fd\u3002\u5176\u4e2d\uff0cgloss \u8868\u57fa \u790e\u7fa9\u539f\u7279\u5fb5 PBFN \u03b1 = \u22120.03 \uff0c\u6700\u597d\u7684\u6548\u80fd\u5230 92.3276%\u3002 92.3276% 92.3276% 95.0% \u5ee3\u7fa9\u77e5\u7db2\u8a5e\u5f59\u610f\ufa0a\u6975\u6027\u7684\u9810\u6e2c 35 5. \u7d50\uf941 \u672c\u7814\u7a76\u4f7f\u7528\uf9ba Google Web 5-gram Version 1 \uf92d\u62bd\u53d6\u8a9e\u7bc7\u7279\u5fb5\uff0c\u4e26\u52a0\u4e0a\uf92d\u81ea E-HowNet \u7684 \u57fa\u790e\u7fa9\u539f\u7279\u5fb5\uff0c\u7528\u76e3\u7763\u5f0f\u6a5f\u5668\u5b78\u7fd2\u7684\u65b9\u6cd5\uff0c\uf92d\u9810\u6e2c E-HowNet \u8a5e\u5f59\u7684\u610f\ufa0a\u6975\u6027\u3002\u96d6\u7136\u55ae \u7368\u4f7f\u7528\uf967\u540c\u7684\u7279\u5fb5\u5df2\u7d93\u53ef\u4ee5\u63a5\u8fd1 90% \u7684\u6e96\u78ba\uf961\uff0c\u4f46\u5982\u679c\u628a\uf978\u7a2e\u7279\u5fb5\u90fd\u52a0\u4ee5\u4f7f\u7528\uff0c\u5206\uf9d0 \u5668\u7684\u6975\u6027\u9810\u6e2c\u7684\u6e96\u78ba\uf961\u53ef\u5230\u9054 92.33% \u7684\u9ad8\u6e96\u78ba\uf961\uff1b\u4ee5\u9019\u7a2e\u65b9\u5f0f\u5efa\uf9f7\u7684\u5206\uf9d0\u5668\uff0c\u53ef\u7528\uf92d \u81ea\u52d5\u6a19\u8a18 E-HowNet \u8a5e\u5f59\u7684\u610f\ufa0a\u6975\u6027\u3002 \u5716 4 \uf9e1\u653f\u5112 \u7b49 \u5716 4. \u5ee3\u7fa9\u77e5\u7db2\u7279\u5fb5\u65bc\uf967\u540c \u03b1 \u503c\u7684\u6548\u80fd\u6bd4\u8f03 \u6211\u5011\u5f9e\u5716 4 \u53ef\u4ee5\u770b\u51fa\uff0c\u63cf\u8ff0 PBFN \u7684\u6298\u7dda\u5728\u6240\u6709\u7684 \u03b1 \u503c\u4e0b\uff0c\u6e96\u78ba\uf961\u7686\uf976\u9ad8\u65bc PBF\uff0c \u4f46\u662f\uf978\u500b\u6700\u5927\u503c (\u03b1 = \u22120.02) \u7684\u5dee\u8ddd\u50c5 0.1724%\uff0c\u6b64\u5dee\u8ddd\u70ba\uf967\u986f\u8457\uff0c\u6aa2\u5b9a\u7d50\u679c (1.50, 0.22)\u3002 \u7531\u65bc \u03b1 < 0 \u6709\u6700\u4f73\u6548\u80fd\uff0c\u9019\u8868\u793a\u6df1\ufa01\u8f03\u6df1\u7d66\u8f03\u9ad8\u6b0a\u91cd\uff0c\u8a72\u7fa9\u539f\u6709\u8f03\u597d\u7684\u7279\u5fb5\uff0c\u53ef\u4ee5\u7d66\u5206 \uf9d0\u5668\u5b78\u7fd2\u3002 4.3 \u8a9e\u7bc7\u7279\u5fb5\u7684\u6548\u80fd \u8a9e\u7bc7\u7279\u5fb5\u4f7f\u7528\u5341\u7d44\u7279\u5fb5\u96c6\u7684\u540d\u7a31\uff0c\u4ee5\u53ca\u7279\u5fb5\uf969\uf97e\uff0c\u5982\u8868 3 \u6240\u793a\u3002\u5728\u8868\u4e2d\uff0c\u6211\u5011\u4f7f\u7528\u7279\u5fb5 \u96c6\u4ee3\u865f\uf92d\u4ee3\u8868\u8a72\u7279\u5fb5\u96c6\u3002\u5341\u7d44\u7279\u5fb5\u96c6\u4e2d\uff0c\u6700\u5c11\u7684\u662f Adj \u7684\u7279\u5fb5\u96c6\uff0c\u53ea\u6709 948 \u500b\u8a5e\uff0c\u6700\u591a \u7684\u662f All \u7684\u7279\u5fb5\u96c6\uff0c\u6709 86,712 \u500b\u8a5e\u3002 \u8868 3. \u8a9e\u7bc7\u7279\u5fb5\u6240\u4f7f\u7528\u7684\u7279\u5fb5\u96c6\u8207\u5176\u7279\u5fb5\uf969 \u7279\u5fb5\u96c6 \u7279\u5fb5\u96c6\u4ee3\u865f \u7279\u5fb5\uf969 \u5ee3\u7fa9\u77e5\u7db2\u540d\u8a5e Noun 46,807 \u5ee3\u7fa9\u77e5\u7db2\u52d5\u8a5e Verb 37,109 \u5ee3\u7fa9\u77e5\u7db2\u526f\u8a5e Adv. 2,364 \u5ee3\u7fa9\u77e5\u7db2\u5f62\u5bb9\u8a5e Adj. 948 All 86,712 \u7d44\u5408\u7279\u5fb5\u6642\uff0c\u56e0\u70ba\u9918\u5f26\u6a19\u6e96\u5316\u6709\u6700\u597d\u7684\u6548\u80fd\uff0c\u6240\u4ee5\u8a9e\u7bc7\u7279\u5fb5\u9078\u64c7\u9918\u5f26\u6a19\u6e96\u5316\u5f8c\u7684\u5341\u7d44\u7279 F5000-2 97.2635% 98.0392% 97.1705% 96.0912% 94.9153% 94.8718% \u7684\uf969\uf97e\u3002 \u5ee3\u7fa9\u77e5\u7db2\u6240\u6709\u8a5e\u5f59 89.0% 89.2% -0.05 -0.04 -0.03 -0.02 -0.01 0 0.01 0.02 0.03 0.04 0.05 Alpha \u6b64\u6548\u80fd\u8ddf\u5176\u4ed6\uf978\u8005\u7684\u5dee\u8ddd\u70ba\u986f\u8457\uff0c\u6aa2\u5b9a\u7d50\u679c (4.61, 0.03)\u3002 \u5716 5. \u4f7f\u7528\u8a9e\u7bc7\u7279\u5fb5\u6642\u7684\u9810\u6e2c\u6548\u80fd \u5716 5 \u4e2d\u7279\u5fb5\u96c6\u7684\u500b\uf969\uff0c\u4e26\u6c92\u6709\u7d55\u5c0d\u7684\u5f71\u97ff\uff0c\u4f46\uf974\u500b\uf969\u592a\u5c11\uff0c\u5982\u7279\u5fb5\u500b\uf969\u5c0f\u65bc 2364 \u500b\uff0c\u5247\u6548\u80fd\u6703\u660e\u986f\u8b8a\u5dee\u3002\u5716 4 \u4e2d\u7684\u6700\u4f73\u503c PBFN(\u03b1 = \u22120.02)\u70ba 89.61%\uff0c\u7279\u5fb5\u500b\uf969\u70ba 2,567 \u500b\uff0c\u9019\u500b\u503c\u6bd4\u5716 5 \u4e2d\u7684\u6700\u4f73\u503c 88.23%\u9084\u8981\u5927\uff0c\u9019\u8868\u793a\u5ee3\u7fa9\u77e5\u7db2\u4e2d\u7684\u7279\u5fb5\u6bd4\u8f03\u6e96\u78ba\uff0c\u4f46\u9019 \u5dee\u8ddd\u70ba\uf967\u986f\u8457\uff0c\u6aa2\u5b9a\u7d50\u679c (2.49, 0.11)\u3002 4.4 \u7d44\u5408\uf967\u540c\u7279\u5fb5\u7684\u6548\u80fd 50.0% 60.0% 70.0% 80.0% 90.0% 100.0% Adj. (948) Adv. (2364) F5000-1 (5000) F5000-2 (5000) F10000-1 (10000) F10000-2 (10000) Verb (37109) NTUSD (42614) Noun (46807) All (86712) Accuracy Feature set Original Frequency Cos-Normalized Frequency Cos-Normalized TFIDF \u5716 6. \u5ee3\u7fa9\u77e5\u7db2\u3001\u8a9e\u7bc7\u7279\u5fb5\u3001\u8207\u7d44\u5408\u7279\u5fb5\u7684\u6e96\u78ba\uf961\u6bd4\u8f03 4.5 \u7d44\u5408\u5f0f\u7684\u76e3\u7763\u5f0f\u6a5f\u5668\u5b78\u7fd2\u6f14\u7b97\u6cd5\u6548\u80fd \u5728\u5716 6 \u4e2d\uff0c\u7d44\u5408\u51fa\u7684\u7279\u5fb5\u96c6\u6709\u5341\u500b\uff0c\u6240\u4ee5\u5171\u6709\u5341\u500b\u5206\uf9d0\u5668\uff0c\u6bcf\u500b\u5206\uf9d0\u5668\u5728\u8a13\uf996\u6642\uff0c\u5c0d\uf967 \u540c\u8a5e\u6027\u6709\uf967\u540c\u7684\u6548\u80fd\uff0c\u6211\u5011\u5c07\u9019\u5341\u500b\u5206\uf9d0\u5668\u5c0d\u65bc\u6bcf\u500b\u8a5e\u6027\u6a19\u8a18\u7684\u6548\u80fd(\u5167\u90e8\u6e2c\u8a66)\u6574\uf9e4 \u6210\u8868 4\u3002 \u8868 4. \u8a13\uf996\u8cc7\uf9be\u96c6\u4e2d\uff0c\u7d44\u5408\u7279\u5fb5\u5c0d\uf967\u540c\u8a5e\u6027\u7684\u6a19\u8a18\u6e96\u78ba\uf961 \u7279\u5fb5\u96c6\u4ee3\u865f \u7e3d\u9ad4\u6548\u80fd \u8a13\uf996\u8cc7\uf9be\u96c6\u4e2d\uff0c\u4f9d\u8a5e\u6027\u5206\u5225\u8a08\u7b97\u7684\u6e96\u78ba\uf961 \u540d\u8a5e \u52d5\u8a5e \u526f\u8a5e \u5f62\u5bb9\u8a5e \u5176\u4ed6 Adj. 94.3223% 95.9559% 94.2167% 89.9023% 93.2203% 82.0513% Adv. 95.3243% 96.5074% 95.2795% 92.1824% 91.5254% 84.6154% F5000-1 96.1000% 97.3039% 96.0110% 92.8339% 94.9153% 89.7436% 75.0% 80.0% 85.0% 90.0% Adj. (948) Adv. (2364) F5000-1 (5000) F5000-2 (5000) F10000-1 (10000) F10000-2 (10000) Verb (37109) NTUSD (42614) Noun (46807) All (86712) Accuracy Feature set Gloss Context Combine gloss and Context \u5167\u90e8\u6e2c\u8a66\u6548\u80fd(inside test)F10000-2+PBFN \u03b1 = \u22120.03 \u7684 97.5005% \u8ddf\u5be6\u969b\u6e2c\u8a66\u6548\u80fd 92. 3276%\u76f8\u6bd4\uff0c\ufa09\u4f4e\uf9ba 5.31%\uff0c\u9019\ufa09\u4f4e\u5e45\ufa01\u4e26\uf967\u5927\uff0c\u986f\u793a\u9019\u5206\uf9d0\u5668\u7684 generalization \u80fd\uf98a\uf967 \u932f\uff0c\u9019\u4e5f\u662f\u4f7f\u7528 Google Web 5-gram \u7684\u512a\u9ede\uff0c\u5b83\u53ef\u7522\u751f\u8f03\u5f37\u5065 (robust) \u7684\u5206\uf9d0\u5668(Bergsma, Pitler, & Lin, 2010)\u3002 \u6211\u5011\u7528\u5167\u90e8\u6e2c\u8a66\u6548\u80fd\uf92d\u6311\u9078\u5206\uf9d0\u5668\uff0c\u4ee5\uf965\u7528\u5728\u7d44\u5408\u5f0f\u7684\u76e3\u7763\u5f0f\u6a5f\u5668\u5b78\u7fd2\u6f14\u7b97\u6cd5\u4e2d\u3002 \u6211\u5011\u5728\u8868 4 \u4e2d\u9078\uf967\u540c\u8a5e\u6027\u505a\u5f97\u6700\u597d\u7684\u5206\uf9d0\u5668\uf92d\u7d44\u5408\uff0c\u5982\u679c\u6548\u80fd\u76f8\u540c\uff0c\u5247\u9078\u7279\u5fb5\uf969\uf97e\u8f03\u5c11 \u7684\u90a3\u4e00\u500b\u5206\uf9d0\u5668\uff0c\u56e0\u70ba\u7279\u5fb5\uf969\u8f03\u5c11\u901a\u5e38\u5728\u672a\u770b\u904e\u7684\u8cc7\uf9be\u96c6\u6703\u505a\u5f97\u8f03\u597d\u3002\u7d44\u5408\u51fa\u7684\u5206\uf9d0\u5668 \u6211\u5011\u7a31\u70ba EnsembleClassifier\uff0c\u5176\u7d50\u679c\uf99c\u5728\u8868 5\uff0c\u5176\u4e2d F10000-2+PBFN \u03b1 = \u22120.03 \u65bc\u5404\u8a5e\u6027\u7684 \u6a19\u8a18\u6548\u80fd\u4e5f\uf99c\u51fa\uf92d\u6bd4\u8f03\u3002 \u8868 5. \u7d44\u5408\u5206\uf9d0\u5668\u65bc\u5404\u8a5e\u6027\u7684\u6a19\u8a18\u6548\u80fd\u53ca\u6bd4\u8f03 \u5206 \uf9d0 \u5668 \u8a5e\u6027 F10000-2+PBFN \u03b1 = \u22120.03 \u5206\uf9d0\u5668 \u65bc\u5404\u8a5e\u6027\u7684\u6a19\u8a18\u6548\u80fd \u7d44\u5408\u5206\uf9d0\u5668 EnsembleClassifier \u65bc\u5404\u8a5e\u6027\u7684\u6a19\u8a18\u6548\u80fd \u6b63\u78ba \u500b\uf969 \u932f\u8aa4 \u500b\uf969 \u6e96\u78ba\uf961 \u4f7f\u7528\u7684 \u5206\uf9d0\u5668 \u6b63\u78ba \u500b\uf969 \u589e\u6e1b \u932f\u8aa4 \u500b\uf969 \u6e96\u78ba\uf961 \u540d\u8a5e 371 37 90.9314% F10000-2 371 (+0) 37 90.9314% \u52d5\u8a5e 1,681 130 92.8216% F10000-2 1,681 (+0) 130 92.8216% \u526f\u8a5e 67 9 88.1579% F5000-2 69 (+2) 7 90.7895% \u5f62\u5bb9\u8a5e 14 1 93.3333% Noun 12 (-2) 3 80.0000% \u5176\u4ed6 9 1 90.0000% F5000-2 9 (+0) 1 90.0000% \u7e3d\uf969 2,142 178 92.3276% 2142 (+0) 178 92.3276% \u8868 5 \u4e2d\uff0c\u6211\u5011\u4e5f\uf99c\u51fa\u6bcf\u7a2e\u8a5e\u6027\u505a\u932f\u8207\u505a\u5c0d\u7684\u500b\uf969\uff0c\u4e26\u4ee5 F10000-2+PBFN \u03b1 = \u22120.03 \u5206\uf9d0 \u5668\u70ba\u57fa\u6e96\uff0c\u770b\u7d44\u5408\u5f8c\u7684\u5206\uf9d0\u5668\uff0c\u5728\u5404\u8a5e\u6027\u4e2d\u505a\u5c0d\u505a\u932f\u7684\u6b21\uf969\u7684\u589e\u6e1b\uff0c\u7528\u62ec\u865f\uf92d\u6a19\u51fa\u589e\u6e1b \u5716 7. \u56db\u7a2e\u65b9\u6cd5\u6548\u80fd\u6bd4\u8f03 \u7531\u65bc\u6211\u5011\u4f7f\u7528 NTUSD\uff0c\u6211\u5011\u60f3\u770b\u770b NTUSD \u4eba\uf9d0\u6a19\u8a18\u7684\u6548\u80fd\u8ddf\u6211\u5011\u5206\uf9d0\u5668\u7684\u6548\u80fd\u6709 \u4f55\u5dee\uf962\u3002\u5728 Ku & Chen (2007)\u7684\u7814\u7a76\u4e2d\uff0c\u5c0d\u8a5e\u6709\u5206\u56db\uf9d0\u6a19\u8a18\uff0c\u5206\u5225\u662f\u6b63\u9762\u3001\u8ca0\u9762\u3001\u4e2d\uf9f7\u3001 \u53ca\u975e\u610f\ufa0a\u8a5e\uff0c\u4e26\u8058\u8acb\u6a19\u8a18\u8005\u5c0d NTUSD \u9032\ufa08\u6a19\u8a18\uff0c\u6211\u5011\u5c07\u8a72\u7814\u7a76\u4e2d\u6a19\u8a18\u8005\u7684\u6700\u4f73\u6a19\u8a18\u6548 \u80fd\u8207\u672c\u7814\u7a76\u7684\u6bd4\u8f03\u5982\u8868 6\u3002\u7531\u65bc\u4eba\uf9d0\u6a19\u8a18\u8005\u662f\u5c07\u8a5e\u5206\u6210\u56db\uf9d0\uff0c\u4f46\u6211\u5011\u7684\u7cfb\u7d71\u53ea\u5206\uf978\uf9d0\uff0c \u6240\u4ee5\u9019\uf969\u64da\u6c92\u6709\u8fa6\u6cd5\u8ddf\u6211\u5011\u7684\u7d50\u679c\u76f4\u63a5\u76f8\u6bd4\u8f03\uff1b\u4f46\u6211\u5011\u4ecd\u53ef\u5f9e\u8868 6 \u4e2d\u770b\u51fa\uff0c\u672c\u7814\u7a76\u6240\u7522 \u751f\u7684\u6b63\u8ca0\u9762\u8a5e\u5f59\u81ea\u52d5\u6a19\u8a18\u6f14\u7b97\u6cd5\uff0c\u5df2\u9054\u5230\uf9ba\u5f88\u9ad8\u7684\u6548\u80fd\u3002 \u8868 6. NTUSD \u6a19\u8a18\u8005\u8207\u672c\u7814\u7a76\u6a19\u8a18\u6548\u80fd\u6bd4\u8f03 \u5206\uf9d0\u5668 Recall Precision F-Measure F10000-2+PBFN \u03b1 = \u22120.03 92.36% 92.20% 92.27% \u4e09\u4eba\u4e2d\u6700\u4f73\u7684\u4eba\uf9d0\u6a19\u8a18\u8005 96.58% 88.87% 92.56% 88.2328% 90.0% Gloss Context Combine gloss and Context Ensemble classifier \u6211\u5011\u5e0c\u671b\u5728\u672a\uf92d\u80fd\u628a\u9019\u7a2e\u65b9\u5f0f\uff0c\u5f80\uf967\u540c\u7684\u65b9\u5411\u64f4\u5c55\uff0c\uf92d\u7d66\u4e88 E-HowNet \u8a5e\u5f59\uf901\u591a\u610f Accuracy \ufa0a\u7684\u5c6c\u6027\uff0c\u9019\u5305\u62ec\u5c0d\u8a5e\u5f59\u6a19\u8a18\u4e3b\u5ba2\u89c0\u7684\u5c6c\u6027\u53ca\u6b63\u8ca0\u9762\u50be\u5411\u7684\u5f37\ufa01\u7b49\u3002\u9664\u6b64\u4e4b\u5916\uff0c\u56e0\u70ba E-HowNet \u8a5e\u5f59\u6709\u8a31\u591a\uf967\u540c\u7684\u8a5e\u6027\uff0c\u6211\u5011\u4e5f\u5e0c\u671b\u80fd\u628a\u6211\u5011\u7684\u65b9\u6cd5\uff0c\u904b\u7528\u8a5e\u6027\u7684\u5c64\u6b21\uf92d\u9032\ufa08 \u6a19\u8a18\u3002\u85c9\u7531\u63d0\u4f9b\uf901\u7cbe\u78ba\u7684\u5b57\u5f59\u610f\ufa0a\u6a19\u8a18\u8cc7\u8a0a\uff0c\uf92d\u652f\u63f4\uf906\u5b50\u53ca\u6587\u4ef6\u5c64\u6b21\u7684\u610f\ufa0a\u5206\u6790\u3002 \u8868 6 89.6121% 85.0% \u81f4\u8b1d</td></tr><tr><td>\u6700\u5e38\u51fa\u73fe 5000 \u8a5e \u5fb5\u96c6\uff0c\u5206\u5225\u8207\u5ee3\u7fa9\u77e5\u7db2\u7279\u5fb5\u6548\u80fd\u6700\u597d\u7684 PBFN \u03b1 = \u22120.03 \u7d44\u5408\uff0c\uf92d\u8a13\uf996\u5206\uf9d0\u5668\uff0c\u5206\uf9d0\u5668\u9810\u6e2c\u6e96 F5000-1 5,000 F10000-1 96.2400% 97.3652% 96.1767% 92.8339% 94.9153% 89.7436% EnsembleClassifier \u6240\u5f97\u6210\u7e3e\u8ddf F10000-2+PBFN \u03b1 = \u22120.03 \u76f8\u540c\uff0c\u9019\u8868\u793a\u76ee\u524d\u7684\u5206\uf9d0\u5668\u7d44</td></tr><tr><td>\u6700\u5e38\u51fa\u73fe 5000 \u8a5e(\u9577\ufa01\u22672) \u78ba\uf961\u5982\u5716 6\u3002\u5176\u4e2d\u5ee3\u7fa9\u77e5\u7db2\u7279\u5fb5\u7684\u7279\u5fb5\u96c6\u6548\u80fd\u70ba\u56fa\u5b9a\uff0c\u56e0\u6b64\u4ee5\u6c34\u5e73\u76f4\u7dda\u8868\u793a(gloss \u90a3\u689d F5000-2 5,000 F10000-2 97.5005% 98.2843% 97.4189% 96.0912% 94.9153% 94.8718% \u5408\u65b9\u5f0f\uff0c\u7121\u6cd5\u63d0\u5347\u6548\u80fd\u3002</td></tr><tr><td>\u6700\u5e38\u51fa\u73fe 10000 \u8a5e \u6298\u7dda)\u3002\u7d44\u5408\u800c\u6210\u7684\u7279\u5fb5\u96c6\uff0c\u4ee5\u300c\u8a9e\u7bc7\u7279\u5fb5\u96c6\u4ee3\u78bc+PBFN \u03b1 = \u22120.03 \u300d\u52a0\u4ee5\u547d\u540d\uff0c\uf9b5\u5982 F10000-1 Verb 96.5632% 97.5490% 96.5079% 94.4625% 91.5254% 89.7436% 10,000 \u300cF10000-2+PBFN \u03b1 = \u22120.03 \u300d\u8868\u793a\u300c\u6700\u5e38\u51fa\u73fe 10000 \u8a5e(\u9577\ufa01\u22672)\u300d\u8ddf\u300cPBFN \u03b1 = \u22120.03 \u300d\uf978 NTUSD 96.8218% 97.3039% 96.8254% 95.1140% 93.2203% 94.8718% \u6700\u5e38\u51fa\u73fe 10000 \u8a5e(\u9577\ufa01\u22672) F10000-2 10,000 \u500b\u7279\u5fb5\u96c6\u7684\u7d44\u5408\u3002 Noun 96.8541% 98.1005% 96.6460% 96.0912% 96.6102% 89.7436%</td></tr><tr><td>NTUSD(\u5b8c\u6574\u7248) \u6211\u5011\u5f9e\u5716 6 \u53ef\u4ee5\u770b\u51fa\uff0c\u5c07\u5ee3\u7fa9\u77e5\u7db2\u7279\u5fb5\u8207\u5916\u90e8\u8a9e\uf9be\u7279\u5fb5\u7d44\u5408\u4e4b\u5f8c\uff0c\u6e96\u78ba\uf961\u90fd\u6709\u986f\u8457 NTUSD 42,614 All 96.4124% 97.4265% 96.3699% 93.1596% 94.9153% 89.7436%</td></tr></table>",
"num": null,
"type_str": "table",
"text": "\u70ba\u57fa\u790e\u7fa9\u539f\u65b9\u6cd5\u5728\uf967\u540c \u03b1 \u503c\u6240\u5f97\u5230\u7684\u9810\u6e2c\u6e96\u78ba\uf961\uff0c\u5176\u4e2d\u516c\u5f0f (2) \u7684\u7d50\u679c\u662f PBF (Prime-Based Feature)\u90a3\u689d\u6298\u7dda\uff0c\u6700\u4f73\u7684 \u03b1 \u503c\u70ba \u22120.02\uff0c\u6e96\u78ba\uf961\u70ba 89.4397%\u3002\u7576 PBF \u4e2d \u03b1 = 0\uff0c\u8a72\u7d50\u679c\u5373\u70ba\u516c\u5f0f (1) \u7684\u7d50\u679c\u3002\u516c\u5f0f (3) \u7684\u7d50\u679c\u662f PBFN(Prime-Based Feature with Negation)\u90a3\u689d\u6298\u7dda\uff0c\u6700\u4f73\u7684 \u03b1 \u503c\u70ba \u22120.02 \u53ca \u22120.03\uff0c\u6e96\u78ba\uf961\u70ba 89.6121%\u3002 \u4e2d\uff0c\u4eba\uf9d0\u6a19\u8a18\u8005\u7684 Recall \u53ca Precision \u53d6\u81ea Ku & Chen (2007)\u3002F10000-2+PBFN \u03b1 = \u22120.03 \u7684\u9810\u6e2c\u7d50\u679c\u70ba (True Positive, False Positive, True Negative, False Negative) = (TP, FP, TN, FN) = (968, 77, 1174, 101)\uff0c\u5176\u4e2d Positive \u8868\u6b63\u9762\u6975\u6027\u3002\u6211\u5011\u5206\u5225\u5c0d\u6b63\u8ca0\u9762\u6975\u6027\u8a08\u7b97 Recall\u3001Precision \u53ca F-Measure (R + \u3001P + \u3001F + \u3001R \u2212 \u3001P \u2212 \u3001F \u2212 )\uff0c\u5176\u4e2d\uff0cP + =TP/(TP+FP)\u3001 R + =TP/(TP+FN)\u3001F + = 2P + R + /(P + +R + )\u3001P \u2212 =TN/(TN+FN)\u3001R \u2212 =TN/(TN+FP)\u3001F \u2212 = 2P \u2212 R \u2212 /(P \u2212 +R \u2212 )\uff0c \u6700\u5f8c\u7cfb\u7d71\u7684 Recall=(R + +R \u2212 )/2\u3001Precision=(P + +P \u2212 )/2 \u53ca F-Measure = (F + +F \u2212 )/2 = (91.58% + 92.95%)/2 = 92.27%\u3002\u7531\u8a08\u7b97\u4e2d\u6211\u5011\u53ef\u4ee5\u770b\u5230\uff0c\u6211\u5011\u7684\u7cfb\u7d71\u5c0d\u8ca0\u9762\u6975\u6027\u505a\u5f97\u8f03\u597d\uff0c\u800c\u4e14\u56e0 \u8cc7\uf9be\u96c6\u6709\u8f03\u591a\u7684\u8ca0\u9762\u8a5e\u5f59\uff0c\u6240\u4ee5\u6700\u5f8c\u7684\u6e96\u78ba\uf961 92.33% \u6bd4 F + \u9ad8\u3002 Research of this paper was partially supported by National Science Council (Taiwan) under the contract NSC 98-2221-E-002-175-MY3.",
"html": null
}
}
}
} |