File size: 50,893 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 | {
"paper_id": "P02-1018",
"header": {
"generated_with": "S2ORC 1.0.0",
"date_generated": "2023-01-19T09:30:45.636243Z"
},
"title": "A simple pattern-matching algorithm for recovering empty nodes and their antecedents *",
"authors": [
{
"first": "Mark",
"middle": [],
"last": "Johnson",
"suffix": "",
"affiliation": {
"laboratory": "Brown Laboratory for Linguistic Information Processing Brown University Mark",
"institution": "",
"location": {}
},
"email": "johnson@brown.edu"
}
],
"year": "",
"venue": null,
"identifiers": {},
"abstract": "This paper describes a simple patternmatching algorithm for recovering empty nodes and identifying their co-indexed antecedents in phrase structure trees that do not contain this information. The patterns are minimal connected tree fragments containing an empty node and all other nodes co-indexed with it. This paper also proposes an evaluation procedure for empty node recovery procedures which is independent of most of the details of phrase structure, which makes it possible to compare the performance of empty node recovery on parser output with the empty node annotations in a goldstandard corpus. Evaluating the algorithm on the output of Charniak's parser (Charniak, 2000) and the Penn treebank (Marcus et al., 1993) shows that the patternmatching algorithm does surprisingly well on the most frequently occuring types of empty nodes given its simplicity. * I would like to thank my colleages in the Brown Laboratory for Linguistic Information Processing (BLLIP) as well as Michael Collins for their advice. This research was supported by NSF awards DMS 0074276 and ITR IIS 0085940. 1 There are other ways to represent this information that do not require empty nodes; however, information about non-local dependencies must be represented somehow in order to interpret these constructions.",
"pdf_parse": {
"paper_id": "P02-1018",
"_pdf_hash": "",
"abstract": [
{
"text": "This paper describes a simple patternmatching algorithm for recovering empty nodes and identifying their co-indexed antecedents in phrase structure trees that do not contain this information. The patterns are minimal connected tree fragments containing an empty node and all other nodes co-indexed with it. This paper also proposes an evaluation procedure for empty node recovery procedures which is independent of most of the details of phrase structure, which makes it possible to compare the performance of empty node recovery on parser output with the empty node annotations in a goldstandard corpus. Evaluating the algorithm on the output of Charniak's parser (Charniak, 2000) and the Penn treebank (Marcus et al., 1993) shows that the patternmatching algorithm does surprisingly well on the most frequently occuring types of empty nodes given its simplicity. * I would like to thank my colleages in the Brown Laboratory for Linguistic Information Processing (BLLIP) as well as Michael Collins for their advice. This research was supported by NSF awards DMS 0074276 and ITR IIS 0085940. 1 There are other ways to represent this information that do not require empty nodes; however, information about non-local dependencies must be represented somehow in order to interpret these constructions.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Abstract",
"sec_num": null
}
],
"body_text": [
{
"text": "One of the main motivations for research on parsing is that syntactic structure provides important information for semantic interpretation; hence syntactic parsing is an important first step in a variety of useful tasks. Broad coverage syntactic parsers with good performance have recently become available (Charniak, 2000; Collins, 2000) , but these typically produce as output a parse tree that only encodes local syntactic information, i.e., a tree that does not include any \"empty nodes\". (Collins (1997) discusses the recovery of one kind of empty node, viz., WH-traces) . This paper describes a simple patternmatching algorithm for post-processing the output of such parsers to add a wide variety of empty nodes to its parse trees.",
"cite_spans": [
{
"start": 307,
"end": 323,
"text": "(Charniak, 2000;",
"ref_id": "BIBREF3"
},
{
"start": 324,
"end": 338,
"text": "Collins, 2000)",
"ref_id": "BIBREF5"
},
{
"start": 493,
"end": 508,
"text": "(Collins (1997)",
"ref_id": "BIBREF4"
},
{
"start": 565,
"end": 575,
"text": "WH-traces)",
"ref_id": null
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "Empty nodes encode additional information about non-local dependencies between words and phrases which is important for the interpretation of constructions such as WH-questions, relative clauses, etc. 1 For example, in the noun phrase the man Sam likes the fact the man is interpreted as the direct object of the verb likes is indicated in Penn treebank notation by empty nodes and coindexation as shown in Figure 1 (see the next section for an explanation of why likes is tagged VBZ t rather than the standard VBZ).",
"cite_spans": [
{
"start": 407,
"end": 413,
"text": "Figure",
"ref_id": null
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "The broad-coverage statistical parsers just mentioned produce a simpler tree structure for such a relative clause that contains neither of the empty nodes just indicated. Rather, they produce trees of the kind shown in Figure 2 . Unlike the tree depicted in Figure 1, this type of tree does not explicitly represent the relationship between likes and the man.",
"cite_spans": [],
"ref_spans": [
{
"start": 219,
"end": 227,
"text": "Figure 2",
"ref_id": "FIGREF1"
},
{
"start": 258,
"end": 264,
"text": "Figure",
"ref_id": null
}
],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "This paper presents an algorithm that takes as its input a tree without empty nodes of the kind shown in Figure 2 and modifies it by inserting empty nodes and coindexation to produce a the tree shown in Figure 1. The algorithm is described in detail in section 2. The standard Parseval precision and recall measures for evaluating parse accuracy do not measure the accuracy of empty node and antecedent recovery, but there is a fairly straightforward extension of them that can evaluate empty node and antecedent recovery, as described in section 3. The rest of this section provides a brief introduction to empty nodes, especially as they are used in the Penn Treebank.",
"cite_spans": [],
"ref_spans": [
{
"start": 105,
"end": 113,
"text": "Figure 2",
"ref_id": "FIGREF1"
},
{
"start": 203,
"end": 209,
"text": "Figure",
"ref_id": null
}
],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "Non-local dependencies and displacement phenomena, such as Passive and WH-movement, have been a central topic of generative linguistics since its inception half a century ago. However, current linguistic research focuses on explaining the possible non-local dependencies, and has little to say about how likely different kinds of dependencies are. Many current linguistic theories of non-local dependencies are extremely complex, and would be difficult to apply with the kind of broad coverage described here. Psycholinguists have also investigated certain kinds of non-local dependencies, and their theories of parsing preferences might serve as the basis for specialized algorithms for recovering certain kinds of non-local dependencies, such as WH dependencies. All of these approaches require considerably more specialized linguitic knowledge than the pattern-matching algorithm described here. This algorithm is both simple and general, and can serve as a benchmark against which more complex approaches can be evaluated. The pattern-matching approach is not tied to any particular linguistic theory, but it does require a treebank training corpus from which the algorithm extracts its patterns. We used sections 2-21 of the Penn Treebank as the training corpus; section 24 was used as the development corpus for experimentation and tuning, while the test corpus (section 23) was used exactly once (to obtain the results in section 3). Chapter 4 of the Penn Treebank tagging guidelines (Bies et al., 1995) contains an extensive description of the kinds of empty nodes and the use of co-indexation in the Penn Treebank. Table 1 contains summary statistics on the distribution of empty nodes in the Penn Treebank. The entry with POS SBAR and no label refers to a \"compound\" type of empty structure labelled SBAR consisting of an empty complementizer and an empty (moved) S (thus SBAR is really a nonterminal label rather than a part of speech); a typical example is shown in Figure 3 . As might be expected the distribution is highly skewed, with most of the empty node tokens belonging to just a few types. Because of this, a system can provide good average performance on all empty nodes if it performs well on the most frequent types of empty nodes, and conversely, a system will perform poorly on average if it does not perform at least moderately well on the most common types of empty nodes, irrespective of how well it performs on more esoteric constructions. Table 1 : The distribution of the 10 most frequent types of empty nodes and their antecedents in sections 2-21 of the Penn Treebank (there are approximately 64,000 empty nodes in total). The \"label\" column gives the terminal label of the empty node, the \"POS\" column gives its preterminal label and the \"Antecedent\" column gives the label of its antecedent. The entry with an SBAR POS and empty label corresponds to an empty compound SBAR subtree, as explained in the text and Figure 3 . be regarded as an instance of the Memory-Based Learning approach, where both the pattern extraction and pattern matching involve recursively visiting all of the subtrees of the tree concerned. It can also be regarded as a kind of tree transformation, so the overall system architecture (including the parser) is an instance of the \"transform-detransform\" approach advocated by Johnson (1998) . The algorithm has two phases. The first phase of the algorithm extracts the patterns from the trees in the training corpus. The second phase of the algorithm uses these extracted patterns to insert empty nodes and index their antecedents in trees that do not contain empty nodes. Before the trees are used in the training and insertion phases they are passed through a common preproccessing step, which relabels preterminal nodes dominating auxiliary verbs and transitive verbs.",
"cite_spans": [
{
"start": 1491,
"end": 1510,
"text": "(Bies et al., 1995)",
"ref_id": "BIBREF2"
},
{
"start": 3335,
"end": 3349,
"text": "Johnson (1998)",
"ref_id": "BIBREF7"
}
],
"ref_spans": [
{
"start": 1624,
"end": 1631,
"text": "Table 1",
"ref_id": null
},
{
"start": 1978,
"end": 1986,
"text": "Figure 3",
"ref_id": "FIGREF2"
},
{
"start": 2470,
"end": 2477,
"text": "Table 1",
"ref_id": null
},
{
"start": 2947,
"end": 2955,
"text": "Figure 3",
"ref_id": "FIGREF2"
}
],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "The preprocessing step relabels auxiliary verbs and transitive verbs in all trees seen by the algorithm. This relabelling is deterministic and depends only on the terminal (i.e., the word) and its preterminal label. Auxiliary verbs such as is and being are relabelled as either a AUX or AUXG respectively. The relabelling of auxiliary verbs was performed primarily because Charniak's parser (which produced one of the test corpora) produces trees with such labels; experiments (on the development section) show that auxiliary relabelling has little effect on the algorithm's performance.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Auxiliary and transitivity annotation",
"sec_num": "2.1"
},
{
"text": "The transitive verb relabelling suffixes the preterminal labels of transitive verbs with \" t\". For example, in Figure 1 the verb likes is relabelled VBZ t in this step. A verb is deemed transitive if its stem is followed by an NP without any grammatical function annotation at least 50% of the time in the training corpus; all such verbs are relabelled whether or not any particular instance is followed by an NP.",
"cite_spans": [],
"ref_spans": [
{
"start": 111,
"end": 119,
"text": "Figure 1",
"ref_id": "FIGREF0"
}
],
"eq_spans": [],
"section": "Auxiliary and transitivity annotation",
"sec_num": "2.1"
},
{
"text": "Intuitively, transitivity would seem to be a powerful cue that there is an empty node following a verb. Experiments on the development corpus showed that transitivity annotation provides a small but useful improvement to the algorithm's performance. The accuracy of transitivity labelling was not systematically evaluated here.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Auxiliary and transitivity annotation",
"sec_num": "2.1"
},
{
"text": "Informally, patterns are minimal connected tree fragments containing an empty node and all nodes co-indexed with it. The intuition is that the path from the empty node to its antecedents specifies important aspects of the context in which the empty node can appear.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Patterns and matchings",
"sec_num": "2.2"
},
{
"text": "There are many different possible ways of realizing this intuition, but all of the ones tried gave approximately similar results so we present the simplest one here. The results given below were generated where the pattern for an empty node is the minimal tree fragment (i.e., connected set of local trees) required to connect the empty node with all of the nodes coindexed with it. Any indices occuring on nodes in the pattern are systematically renumbered beginning with 1. If an empty node does not bear an index, its pattern is just the local tree containing it. Figure 4 displays the single pattern that would be extracted corresponding to the two empty nodes in the tree depicted in Figure 1 .",
"cite_spans": [],
"ref_spans": [
{
"start": 567,
"end": 575,
"text": "Figure 4",
"ref_id": "FIGREF3"
},
{
"start": 689,
"end": 697,
"text": "Figure 1",
"ref_id": "FIGREF0"
}
],
"eq_spans": [],
"section": "Patterns and matchings",
"sec_num": "2.2"
},
{
"text": "For this kind of pattern we define pattern matching informally as follows. If p is a pattern and t is a tree, then p matches t iff t is an extension of p ignoring empty nodes in p. For example, the pattern displayed in Figure 4 matches the subtree rooted under SBAR depicted in Figure 2 .",
"cite_spans": [],
"ref_spans": [
{
"start": 219,
"end": 227,
"text": "Figure 4",
"ref_id": "FIGREF3"
},
{
"start": 278,
"end": 286,
"text": "Figure 2",
"ref_id": "FIGREF1"
}
],
"eq_spans": [],
"section": "Patterns and matchings",
"sec_num": "2.2"
},
{
"text": "If a pattern p matches a tree t, then it is possible to substitute p for the fragment of t that it matches. For example, the result of substituting the pattern shown in Figure 4 for the subtree rooted under SBAR depicted in Figure 2 is the tree shown in Figure 1 . Note that the substitution process must \"standardize apart\" or renumber indices appropriately in order to avoid accidentally labelling empty nodes inserted by two independent patterns with the same index.",
"cite_spans": [],
"ref_spans": [
{
"start": 169,
"end": 177,
"text": "Figure 4",
"ref_id": "FIGREF3"
},
{
"start": 224,
"end": 232,
"text": "Figure 2",
"ref_id": "FIGREF1"
},
{
"start": 254,
"end": 262,
"text": "Figure 1",
"ref_id": "FIGREF0"
}
],
"eq_spans": [],
"section": "Patterns and matchings",
"sec_num": "2.2"
},
{
"text": "Pattern matching and substitution can be defined more rigorously using tree automata (G\u00e9cseg and Steinby, 1984) , but for reasons of space these definitions are not given here.",
"cite_spans": [
{
"start": 85,
"end": 111,
"text": "(G\u00e9cseg and Steinby, 1984)",
"ref_id": "BIBREF6"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Patterns and matchings",
"sec_num": "2.2"
},
{
"text": "In fact, the actual implementation of pattern matching and substitution used here is considerably more complex than just described. It goes to some lengths to handle complex cases such as adjunction and where two or more empty nodes' paths cross (in these cases the pattern extracted consists of the union of the local trees that constitute the patterns for each of the empty nodes). However, given the low frequency of these constructions, there is probably only one case where this extra complexity is justified: viz., the empty compound SBAR subtree shown in Figure 3 .",
"cite_spans": [],
"ref_spans": [
{
"start": 562,
"end": 570,
"text": "Figure 3",
"ref_id": "FIGREF2"
}
],
"eq_spans": [],
"section": "Patterns and matchings",
"sec_num": "2.2"
},
{
"text": "Suppose we have a rank-ordered list of patterns (the next subsection describes how to obtain such a list). The procedure that uses these to insert empty nodes into a tree t not containing empty nodes is as follows. We perform a pre-order traversal of the subtrees of t (i.e., visit parents before their children), and at each subtree we find the set of patterns that match the subtree. If this set is non-empty we substitute the highest ranked pattern in the set into the subtree, inserting an empty node and (if required) co-indexing it with its antecedents.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Empty node insertion",
"sec_num": "2.3"
},
{
"text": "Note that the use of a pre-order traversal effectively biases the procedure toward \"deeper\", more embedded patterns. Since empty nodes are typically located in the most embedded local trees of patterns (i.e., movement is usually \"upward\" in a tree), if two different patterns (corresponding to different non-local dependencies) could potentially insert empty nodes into the same tree fragment in t, the deeper pattern will match at a higher node in t, and hence will be substituted. Since the substitution of one pattern typically destroys the context for a match of another pattern, the shallower patterns no longer match. On the other hand, since shal-lower patterns contain less structure they are likely to match a greater variety of trees than the deeper patterns, they still have ample opportunity to apply.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Empty node insertion",
"sec_num": "2.3"
},
{
"text": "Finally, the pattern matching process can be speeded considerably by indexing patterns appropriately, since the number of patterns involved is quite large (approximately 11,000). For patterns of the kind described here, patterns can be indexed on their topmost local tree (i.e., the pattern's root node label and the sequence of node labels of its children).",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Empty node insertion",
"sec_num": "2.3"
},
{
"text": "After relabelling preterminals as described above, patterns are extracted during a traversal of each of the trees in the training corpus. Table 2 lists the most frequent patterns extracted from the Penn Treebank training corpus. The algorithm also records how often each pattern was seen; this is shown in the \"count\" column of Table 2 .",
"cite_spans": [],
"ref_spans": [
{
"start": 138,
"end": 145,
"text": "Table 2",
"ref_id": null
},
{
"start": 328,
"end": 335,
"text": "Table 2",
"ref_id": null
}
],
"eq_spans": [],
"section": "Pattern extraction",
"sec_num": "2.4"
},
{
"text": "The next step of the algorithm determines approximately how many times each pattern can match some subtree of a version of the training corpus from which all empty nodes have been removed (regardless of whether or not the corresponding substitutions would insert empty nodes correctly). This information is shown under the \"match\" column in Table 2, and is used to filter patterns which would most often be incorrect to apply even though they match. If c is the count value for a pattern and m is its match value, then the algorithm discards that pattern when the lower bound of a 67% confidence interval for its success probability (given c successes out of m trials) is less than 1/2. This is a standard technique for \"discounting\" success probabilities from small sample size data (Witten and Frank, 2000) . (As explained immediately below, the estimates of c and m given in Table 2 are inaccurate, so whenever the estimate of m is less than c we replace m by c in this calculation). This pruning removes approximately 2,000 patterns, leaving 9,000 patterns.",
"cite_spans": [
{
"start": 784,
"end": 808,
"text": "(Witten and Frank, 2000)",
"ref_id": "BIBREF9"
}
],
"ref_spans": [
{
"start": 878,
"end": 885,
"text": "Table 2",
"ref_id": null
}
],
"eq_spans": [],
"section": "Pattern extraction",
"sec_num": "2.4"
},
{
"text": "The match value is obtained by making a second pre-order traversal through a version of the training data from which empty nodes are removed. It turns out that subtle differences in how the match value is obtained make a large difference to the algorithm's performance. Initially we defined the match value of a pattern to be the number of subtrees that match that pattern in the training corpus. But as ex-plained above, the earlier substitution of a deeper pattern may prevent smaller patterns from applying, so this simple definition of match value undoubtedly over-estimates the number of times shallow patterns might apply. To avoid this over-estimation, after we have matched all patterns against a node of a training corpus tree we determine the correct pattern (if any) to apply in order to recover the empty nodes that were originally present, and reinsert the relevant empty nodes. This blocks the matching of shallower patterns, reducing their match values and hence raising their success probability. (Undoubtedly the \"count\" values are also over-estimated in the same way; however, experiments showed that estimating count values in a similar manner to the way in which match values are estimated reduces the algorithm's performance).",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Pattern extraction",
"sec_num": "2.4"
},
{
"text": "Finally, we rank all of the remaining patterns. We experimented with several different ranking criteria, including pattern depth, success probability (i.e., c/m) and discounted success probability. Perhaps surprisingly, all produced similiar results on the development corpus. We used pattern depth as the ranking criterion to produce the results reported below because it ensures that \"deep\" patterns receive a chance to apply. For example, this ensures that the pattern inserting an empty NP * and WHNP can apply before the pattern inserting an empty complementizer 0.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Pattern extraction",
"sec_num": "2.4"
},
{
"text": "The previous section described an algorithm for restoring empty nodes and co-indexing their antecedents. This section describes two evaluation procedures for such algorithms. The first, which measures the accuracy of empty node recovery but not co-indexation, is just the standard Parseval evaluation applied to empty nodes only, viz., precision and recall and scores derived from these. In this evaluation, each node is represented by a triple consisting of its category and its left and right string positions. (Note that because empty nodes dominate the empty string, their left and right string positions of empty nodes are always identical).",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Empty node recovery evaluation",
"sec_num": "3"
},
{
"text": "Let G be the set of such empty node representations derived from the \"gold standard\" evaluation corpus and T the set of empty node representations Table 2 : The most common empty node patterns found in the Penn Treebank training corpus. The Count column is the number of times the pattern was found, and the Match column is an estimate of the number of times that this pattern matches some subtree in the training corpus during empty node recovery, as explained in the text.",
"cite_spans": [],
"ref_spans": [
{
"start": 147,
"end": 154,
"text": "Table 2",
"ref_id": null
}
],
"eq_spans": [],
"section": "Empty node recovery evaluation",
"sec_num": "3"
},
{
"text": "derived from the corpus to be evaluated. Then as is standard, the precision P , recall R and f-score f are calculated as follows: Table 3 provides these measures for two different test corpora: (i) a version of section 23 of the Penn Treebank from which empty nodes, indices and unary branching chains consisting of nodes of the same category were removed, and (ii) the trees produced by Charniak's parser on the strings of section 23 (Charniak, 2000) . To evaluate co-indexation of empty nodes and their antecedents, we augment the representation of empty nodes as follows. The augmented representation for empty nodes consists of the triple of category plus string positions as above, together with the set of triples of all of the non-empty nodes the empty node is co-indexed with. (Usually this set of antecedents is either empty or contains a single node). Precision, recall and f-score are defined for these augmented representations as before.",
"cite_spans": [
{
"start": 435,
"end": 451,
"text": "(Charniak, 2000)",
"ref_id": "BIBREF3"
}
],
"ref_spans": [
{
"start": 130,
"end": 137,
"text": "Table 3",
"ref_id": null
}
],
"eq_spans": [],
"section": "Empty node recovery evaluation",
"sec_num": "3"
},
{
"text": "P = |G \u2229 T | |T | R = |G \u2229 T | |G| f = 2 P R P + R",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Empty node recovery evaluation",
"sec_num": "3"
},
{
"text": "Note that this is a particularly stringent evaluation measure for a system including a parser, since it is necessary for the parser to produce a non-empty node of the correct category in the correct location to serve as an antecedent for the empty node. Table 4 provides these measures for the same two corpora described earlier.",
"cite_spans": [],
"ref_spans": [
{
"start": 254,
"end": 261,
"text": "Table 4",
"ref_id": null
}
],
"eq_spans": [],
"section": "Empty node recovery evaluation",
"sec_num": "3"
},
{
"text": "In an attempt to devise an evaluation measure for empty node co-indexation that depends less on syntactic structure we experimented with a modified augmented empty node representation in which each antecedent is represented by its head's category and location. (The intuition behind this is that we do not want to penalize the empty node antecedentfinding algorithm if the parser misattaches modifiers to the antecedent). In fact this head-based antecedent representation yields scores very similiar to those obtained using the phrase-based representation. It seems that in the cases where the parser does not construct a phrase in the appropriate location to serve as the antecedent for an empty node, the syntactic structure is typically so distorted that either the pattern-matcher fails or the head-finding algorithm does not return the \"correct\" head either. Table 3 : Evaluation of the empty node restoration procedure ignoring antecedents. Individual results are reported for all types of empty node that occured more than 100 times in the \"gold standard\" corpus (section 23 of the Penn Treebank); these are ordered by frequency of occurence in the gold standard. Section 23 is a test corpus consisting of a version of section 23 from which all empty nodes and indices were removed. The parser output was produced by Charniak's parser (Charniak, 2000) . Table 4 : Evaluation of the empty node restoration procedure including antecedent indexing, using the measure explained in the text. Other details are the same as in Table 4 .",
"cite_spans": [
{
"start": 1342,
"end": 1358,
"text": "(Charniak, 2000)",
"ref_id": "BIBREF3"
}
],
"ref_spans": [
{
"start": 864,
"end": 871,
"text": "Table 3",
"ref_id": null
},
{
"start": 1361,
"end": 1368,
"text": "Table 4",
"ref_id": null
},
{
"start": 1527,
"end": 1534,
"text": "Table 4",
"ref_id": null
}
],
"eq_spans": [],
"section": "Empty node recovery evaluation",
"sec_num": "3"
},
{
"text": "This paper described a simple pattern-matching algorithm for restoring empty nodes in parse trees that do not contain them, and appropriately indexing these nodes with their antecedents. The patternmatching algorithm combines both simplicity and reasonable performance over the frequently occuring types of empty nodes. Performance drops considerably when using trees produced by the parser, even though this parser's precision and recall is around 0.9. Presumably this is because the pattern matching technique requires that the parser correctly identify large tree fragments that encode long-range dependencies not captured by the parser. If the parser makes a single parsing error anywhere in the tree fragment matched by a pattern, the pattern will no longer match. This is not unlikely since the statistical model used by the parser does not model these larger tree fragments. It suggests that one might improve performance by integrating parsing, empty node recovery and antecedent finding in a single system, in which case the current algorithm might serve as a useful baseline. Alternatively, one might try to design a \"sloppy\" pattern matching algorithm which in effect recognizes and corrects common parser errors in these constructions. Also, it is undoubtedly possible to build programs that can do better than this algorithm on special cases. For example, we constructed a Boosting classifier which does recover *U* and empty complementizers 0 more accurately than the pattern-matcher described here (although the pattern-matching algorithm does quite well on these constructions), but this classifier's performance averaged over all empty node types was approximately the same as the pattern-matching algorithm.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Conclusion",
"sec_num": "4"
},
{
"text": "As a comparison of tables 3 and 4 shows, the pattern-matching algorithm's biggest weakness is its inability to correctly distinguish co-indexed NP * (i.e., NP PRO) from free (i.e., unindexed) NP *. This seems to be a hard problem, and lexical information (especially the class of the governing verb) seems relevant. We experimented with specialized classifiers for determining if an NP * is co-indexed, but they did not perform much better than the algorithm presented here. (Also, while we did not sys-tematically investigate this, there seems to be a number of errors in the annotation of free vs. co-indexed NP * in the treebank).",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Conclusion",
"sec_num": "4"
},
{
"text": "There are modications and variations on this algorithm that are worth exploring in future work. We experimented with lexicalizing patterns, but the simple method we tried did not improve results. Inspired by results suggesting that the patternmatching algorithm suffers from over-learning (e.g., testing on the training corpus), we experimented with more abstract \"skeletal\" patterns, which improved performance on some types of empty nodes but hurt performance on others, leaving overall performance approximately unchanged. Possibly there is a way to use both skeletal and the original kind of patterns in a single system.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Conclusion",
"sec_num": "4"
}
],
"back_matter": [],
"bib_entries": {
"BIBREF2": {
"ref_id": "b2",
"title": "Bracketting Guideliness for Treebank II style Penn Treebank Project. Linguistic Data Consortium",
"authors": [
{
"first": "Ann",
"middle": [],
"last": "Bies",
"suffix": ""
},
{
"first": "Mark",
"middle": [],
"last": "Ferguson",
"suffix": ""
},
{
"first": "Karen",
"middle": [],
"last": "Katz",
"suffix": ""
},
{
"first": "Robert",
"middle": [],
"last": "Mac-Intyre",
"suffix": ""
}
],
"year": 1995,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Ann Bies, Mark Ferguson, Karen Katz, and Robert Mac- Intyre, 1995. Bracketting Guideliness for Treebank II style Penn Treebank Project. Linguistic Data Consor- tium.",
"links": null
},
"BIBREF3": {
"ref_id": "b3",
"title": "A maximum-entropy-inspired parser",
"authors": [
{
"first": "Eugene",
"middle": [],
"last": "Charniak",
"suffix": ""
}
],
"year": 2000,
"venue": "The Proceedings of the North American Chapter of the Association for Computational Linguistics",
"volume": "",
"issue": "",
"pages": "132--139",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Eugene Charniak. 2000. A maximum-entropy-inspired parser. In The Proceedings of the North American Chapter of the Association for Computational Linguis- tics, pages 132-139.",
"links": null
},
"BIBREF4": {
"ref_id": "b4",
"title": "Three generative, lexicalised models for statistical parsing",
"authors": [
{
"first": "Michael",
"middle": [],
"last": "Collins",
"suffix": ""
}
],
"year": 1997,
"venue": "The Proceedings of the 35th Annual Meeting of the Association for Computational Linguistics",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Michael Collins. 1997. Three generative, lexicalised models for statistical parsing. In The Proceedings of the 35th Annual Meeting of the Association for Com- putational Linguistics, San Francisco. Morgan Kauf- mann.",
"links": null
},
"BIBREF5": {
"ref_id": "b5",
"title": "Discriminative reranking for natural language parsing",
"authors": [
{
"first": "Michael",
"middle": [],
"last": "Collins",
"suffix": ""
}
],
"year": 2000,
"venue": "Machine Learning: Proceedings of the Seventeenth International Conference (ICML 2000)",
"volume": "",
"issue": "",
"pages": "175--182",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Michael Collins. 2000. Discriminative reranking for nat- ural language parsing. In Machine Learning: Pro- ceedings of the Seventeenth International Conference (ICML 2000), pages 175-182, Stanford, California.",
"links": null
},
"BIBREF6": {
"ref_id": "b6",
"title": "Tree Automata",
"authors": [
{
"first": "Ferenc",
"middle": [],
"last": "G\u00e9cseg",
"suffix": ""
},
{
"first": "Magnus",
"middle": [],
"last": "Steinby",
"suffix": ""
}
],
"year": 1984,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Ferenc G\u00e9cseg and Magnus Steinby. 1984. Tree Au- tomata. Akad\u00e9miai Kiad\u00f3, Budapest.",
"links": null
},
"BIBREF7": {
"ref_id": "b7",
"title": "PCFG models of linguistic tree representations",
"authors": [
{
"first": "Mark",
"middle": [],
"last": "Johnson",
"suffix": ""
}
],
"year": 1998,
"venue": "Computational Linguistics",
"volume": "24",
"issue": "4",
"pages": "613--632",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Mark Johnson. 1998. PCFG models of linguis- tic tree representations. Computational Linguistics, 24(4):613-632.",
"links": null
},
"BIBREF8": {
"ref_id": "b8",
"title": "Building a large annotated corpus of English: The Penn Treebank",
"authors": [
{
"first": "Michell",
"middle": [
"P"
],
"last": "Marcus",
"suffix": ""
},
{
"first": "Beatrice",
"middle": [],
"last": "Santorini",
"suffix": ""
},
{
"first": "Mary",
"middle": [
"Ann"
],
"last": "Marcinkiewicz",
"suffix": ""
}
],
"year": 1993,
"venue": "Computational Linguistics",
"volume": "19",
"issue": "2",
"pages": "313--330",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Michell P. Marcus, Beatrice Santorini, and Mary Ann Marcinkiewicz. 1993. Building a large annotated cor- pus of English: The Penn Treebank. Computational Linguistics, 19(2):313-330.",
"links": null
},
"BIBREF9": {
"ref_id": "b9",
"title": "Data mining: practical machine learning tools and techniques with Java implementations",
"authors": [
{
"first": "H",
"middle": [],
"last": "Ian",
"suffix": ""
},
{
"first": "Eibe",
"middle": [],
"last": "Witten",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Frank",
"suffix": ""
}
],
"year": 2000,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Ian H. Witten and Eibe Frank. 2000. Data mining: prac- tical machine learning tools and techniques with Java implementations. Morgan Kaufmann, San Francisco.",
"links": null
}
},
"ref_entries": {
"FIGREF0": {
"type_str": "figure",
"uris": null,
"num": null,
"text": "A tree containing empty nodes."
},
"FIGREF1": {
"type_str": "figure",
"uris": null,
"num": null,
"text": "A typical parse tree produced by broadcoverage statistical parser lacking empty nodes."
},
"FIGREF2": {
"type_str": "figure",
"uris": null,
"num": null,
"text": "A parse tree containing an empty compound SBAR subtree."
},
"FIGREF3": {
"type_str": "figure",
"uris": null,
"num": null,
"text": "A pattern extracted from the tree displayed inFigure 1."
}
}
}
} |