File size: 38,284 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 |
{
"paper_id": "W98-0115",
"header": {
"generated_with": "S2ORC 1.0.0",
"date_generated": "2023-01-19T06:03:34.539335Z"
},
"title": "Towards a Workbench for Schema-TAGs*",
"authors": [
{
"first": "Karin",
"middle": [],
"last": "Harbusch",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "University of Koblenz-Landau",
"location": {}
},
"email": ""
},
{
"first": "Friedbert",
"middle": [],
"last": "Widmann",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "University of Koblenz-Landau",
"location": {}
},
"email": ""
},
{
"first": "Jens",
"middle": [],
"last": "Woch",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "University of Koblenz-Landau",
"location": {}
},
"email": ""
}
],
"year": "",
"venue": null,
"identifiers": {},
"abstract": "In the following the components of a workbench for the grommar formalism of Schema-Tree Adjoining Grammars (S-TAGs) are outlined. This workbench can also serve as a workbench for pure TA Gs because it provides a component which transforms an arbitrory TAG into an STAG in a non-trivial manner. Another interesting property of the workbench is that it provides a parser, which is realized as a reversible component to generote as weil.",
"pdf_parse": {
"paper_id": "W98-0115",
"_pdf_hash": "",
"abstract": [
{
"text": "In the following the components of a workbench for the grommar formalism of Schema-Tree Adjoining Grammars (S-TAGs) are outlined. This workbench can also serve as a workbench for pure TA Gs because it provides a component which transforms an arbitrory TAG into an STAG in a non-trivial manner. Another interesting property of the workbench is that it provides a parser, which is realized as a reversible component to generote as weil.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Abstract",
"sec_num": null
}
],
"body_text": [
{
"text": "The formalism of augmenting Tree Adjoining Grommars with schemata was introduced in [Weir 87) in order to compress syntactic descriptions. For that purpose, a TAG (see, e.g., (Joshi 86] ) is extended in order to provide the facility to specify a regular expression \u2022\u2022 (RE). A RE is of type a.b, a+b, a+, a\u2022 and a<OJn) 1 where a, b can uniquely refer to child nodes (via Gorn numbers) or a tree-modifying reference of the form g 1 -g21 where g1, g2 are Gorn numbers and g 2 denotes a subtree of g1. This expression means that the subtree g, in gi is ignored and replaced with E. Finally, a,b can be regular expressions themselves. Regular expressions are annotated at each inner node of an elementary tree. The resulting tree is called a schematic elementary tree. Such a tree denotes an elementary tree set just as a regular expression denotes some regular set. Thus, an individual scheme corresponds to a -possibly infinite -set of elementary trees, but itself is not the structural element to build derivation trees of.",
"cite_spans": [
{
"start": 175,
"end": 185,
"text": "(Joshi 86]",
"ref_id": null
}
],
"ref_spans": [],
"eq_spans": [],
"section": "lnt rod uction",
"sec_num": null
},
{
"text": "In order to stress the power of compressing a ~am mar let us reconsider the coordination constru~tio~ proposed in (Weir 87) . In Fig. 1 , the root node NP of the substitution tree t 1 (which is element in the set of initial trees I) is annotated with a regular expression. In this regular expression, the Garn number lnl refers to the",
"cite_spans": [
{
"start": 114,
"end": 123,
"text": "(Weir 87)",
"ref_id": null
}
],
"ref_spans": [
{
"start": 129,
"end": 135,
"text": "Fig. 1",
"ref_id": null
}
],
"eq_spans": [],
"section": "lnt rod uction",
"sec_num": null
},
{
"text": "\u2022Tbis work is partially funded by the DFG -German Research Foundatioo -uoder grant HA 2716/1-1. 58 n-th daughter of the node. For an illustration of this reference in the figure the numbers are explicitely annotated to the individual nodes. For instance, the regular Figure 1 : Coordination of NPs expression 121 at the node NP in t 1 r~presents the tree with the root NP and the unique daughter N -e.g., producing \"John\". The operation \".\" concatenates siblings in the same currently evaluated elementary tree. Accordingly, lll.J21 produces an elementary tree where DET and N are the two daughters of NP ( \"a man\"). The operation \"+\" enumerates alternative elementary trees. For instance, the regular expression 121 + Jll.121 enumerates the two trees mentioned above. The expo-",
"cite_spans": [],
"ref_spans": [
{
"start": 267,
"end": 275,
"text": "Figure 1",
"ref_id": null
}
],
"eq_spans": [],
"section": "lnt rod uction",
"sec_num": null
},
{
"text": "~121+(121+111.121)+.131.021+111.121) DET N CONJ 2 3 el-t1 : NP N N N Bob Bill Mcry Sue end N the dog",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "lnt rod uction",
"sec_num": null
},
{
"text": "nent \"+\" d \" \" d . fi . s",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "lnt rod uction",
"sec_num": null
},
{
"text": "an * pro uce m mte sets of elementary . trees where the construction marked with such an exponent can be repeated arbitrarily often (\"+\" represents the infinite repetition exclusing zero occurrences and \"*\" indusing zero). For instance, tt can produce \"Bob Bill Mary Sue and the dog\" (see tree el-tl in Fig. 1) but not \"and the dog\" because (121 + lll.121)+ prevents the zero repetition so that at least N occurs. Furthermore a single \"and\" cannot be produced because no alternative in the regular expression at the root node starts with 131. A finite number of repetitions can be written with the exponent jxfUlkJ, where the component with the Gorn number x occurs at least l and up to k times.",
"cite_spans": [],
"ref_spans": [
{
"start": 303,
"end": 310,
"text": "Fig. 1)",
"ref_id": null
}
],
"eq_spans": [],
"section": "lnt rod uction",
"sec_num": null
},
{
"text": "Note, that the example is not lexicalized because Weir's dissertation proposal was earlier published than the definition of lexicalization (cf. (Schabes 90)). The coordination with Schema-TAGs works similarly with lexicalization. Accordingly, the root node ha.5 two children (Simple.NP..!-and CONJ) and the RE is \"lll + (111+ .121.JII)\". The substitution tree Simple.NP has two children (DET and N) ",
"cite_spans": [
{
"start": 387,
"end": 398,
"text": "(DET and N)",
"ref_id": null
}
],
"ref_spans": [],
"eq_spans": [],
"section": "lnt rod uction",
"sec_num": null
},
{
"text": "With respect to lexicalized TAGs [Schabes 90)) where each tree in the set of initial and auxiliary trees has at least one lexical leaf {called anchor) no lexicon component is required (cf. XTAG [Daran et al. 94] ). But since the workbench should not determine the grammar formalism it is possible to specify a non-lexicalized TAG ag well.",
"cite_spans": [
{
"start": 189,
"end": 211,
"text": "XTAG [Daran et al. 94]",
"ref_id": null
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Writing Grammar and Lexicon Rules",
"sec_num": null
},
{
"text": "A main emphagis lies on the facility to transform an arbitrary TAG into an STAG. Obviously, an arbitrary TAG G can trivially be transformed into an S-TAG G' by annotating the concatenation of all daughters from left to right at each inner node of each elementary tree. Obviously, this transformation involves no compression. Therefore, the transformation component of the STAGWB produces an S-TAG which guarantees that each label at the root node occurs only once in the set of initial and auxiliary trees.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Writing Grammar and Lexicon Rules",
"sec_num": null
},
{
"text": "The component pedorms the following steps. Firstly, in all elementary trees all subtrees which do not contain the foot node are rewritten by substitution in order to find shared structures 1 \u2022 Since new non-terminals must be introduced to prevent the grammar from overgeneration, the adjoinable auxiliary trees are duplicated and root and foot nodes are renamed by the new nonterrninals. Now, all alternatives for the same root node are collected. For each elementary tree where the root node is labelled with X (b 1 , \u2022\u2022\u2022 , bn), a new schematic tree sx is introduced to the S-TAG G' where its root node is labelled with X and the children result from enumerating all occurring children in all elementary trees bli \"., b 0 without repeating the same label. In the ((((\"NP\" . \"\"))) (({(\"DetP\" . \"\")) :substp T)) ((((\" N\" . \"\")) :headp T))) {{((\"NP\".\"\"))) ((((\"N\". \"\")) :headp T))) ((((\"NP\" . \"r\"))) ((((\"N\" . \"\")) :headp T)) (l(\\ 'S\" . \"\" )) :substp T))) (({{\"NP\".\"\"))) {(((\"DetP\". \"\")) :substp T)) ((((\"N\". \"r\")):constraints \"NA\" :constrainMype :NA) ((((\"N\" . \"\")) :headp T)) ((((\"S\" . \"\")) :substp T)))) ((((\"NP\" . \"\"))) {(((\"G\" . '\"')) :headp T))) ((((\"NP\" . \"g''))) ((((\"NP\" . \"\")) :substp T)) ({((\"G\" . \"'')) :headp T)))",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Writing Grammar and Lexicon Rules",
"sec_num": null
},
{
"text": ".J.l An instance of a grammar transformation is shown in Fig. 2 2 \u2022 Note, that here the first step of introducing substitutions does not have to do much, because most lexicalized TAGs already use substitution. The only new substitution node is N\u00b0.",
"cite_spans": [],
"ref_spans": [
{
"start": 57,
"end": 65,
"text": "Fig. 2 2",
"ref_id": "FIGREF0"
}
],
"eq_spans": [],
"section": "Writing Grammar and Lexicon Rules",
"sec_num": null
},
{
"text": "The resulting REs can be reformulated applying the following transformation rules:",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Writing Grammar and Lexicon Rules",
"sec_num": null
},
{
"text": "1. O')'(llk} .')'\u00df = O')'(llk+l) \u00df, 2. a('Y.<5i)\u00df + \". + a('Y.om)\u00df = 0')'.(01 + \". + Om)\u00df 3. O\"'f\u00df +a\u00df= cry(Olll\u00df where o, \u00df, ')', 0 1 , .\", Om are arbitrary complex REs.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Writing Grammar and Lexicon Rules",
"sec_num": null
},
{
"text": "Note, that different compressing strategies result in different REs. For analysis grammars the rule of factoring out common prefixes is convenient, whereag the factorization according to common hcads is more adequate in generation. E.g. in the example in Fig. 2 for analysis the two alternatives lll.121 and IIl-141 result in IIl.(121 + 141). For generation the alternatives lll.121 + ]21 +!21-131 result in ll l(OJI) .121+121.131. Additionally, this example illustrates that an LD/LP-Schema-TAG can be advantageous especially for generation because there the alternative !2!.131 can easily be incorporated in the compact expression. Now, the automatically introduced substitution trees can be replaced with their original substructures and furthermore all added auxiliary trees can be eliminated again if desired. So the graaunar becomes as lexicalized as it was before. Finally, in order to introduce cu\u2022 to the annotations the following process is carried out. According to the annotation of each substitution node r substitution trees s 1 and s2 are identified which only ' differ in one leaf l in s 1 \u2022 For these candidates the structure must match beside the path to l. If so, the substitution of tree s 1 is explicitely realized and r is modified to refer to s 1 -<path-to-l> instead of referring to s2.",
"cite_spans": [],
"ref_spans": [
{
"start": 255,
"end": 261,
"text": "Fig. 2",
"ref_id": "FIGREF0"
}
],
"eq_spans": [],
"section": "Writing Grammar and Lexicon Rules",
"sec_num": null
},
{
"text": "Tobe able to deal with REs and substitutions the parser extends the Earley-based TAG-parser by [Schabes 90] as follows:",
"cite_spans": [
{
"start": 95,
"end": 107,
"text": "[Schabes 90]",
"ref_id": null
}
],
"ref_spans": [],
"eq_spans": [],
"section": "S-TAG Parser",
"sec_num": null
},
{
"text": "Instead of computing the set of trees described by schemata (which is impossible due to its infinity) explicitely, the REs are interpreted as follows (cf. [Harbusch 94 )): To indicate a certain position, 0 is used to point into the current RE, i.e. a: 0 \u00df indicates, that a: already has been computed. Then, two functions are introduced, namely SHIFT(t/J), which shifts 0 to the right, a.nd NEXT(t/i), which returns a set of nodes to be computed next. SHIFT is performed in each parsing step, in which the computation of a certain node is completed (indicated by raising the dot position to \"ra\"): scanning of terminals (scanner), the prediction of the right part 3 of auxiliary trees (right prediction) in which no prediction to\u00f6k place, and the completion of a root node of a.n auxiliary tree (right completion).",
"cite_spans": [
{
"start": 155,
"end": 167,
"text": "[Harbusch 94",
"ref_id": null
}
],
"ref_spans": [],
"eq_spans": [],
"section": "S-TAG Parser",
"sec_num": null
},
{
"text": "The output of NEXT is responsible for the computation of all alternatives given in the currently considered RE. Thus, each alternative g in \u00df of NEXT(a: 0 \u00df) has to be taken into account for the prediction of new items. This is done in move dot dovn. Whenever a.n elimination ja -bl occurs, it is deferred until node b is actually computed. Instead of processing b an f-scan is simulated. This usually is done in scan obviously, but also may take place in left prediction, if b is non-terminal.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "S-TAG Parser",
"sec_num": null
},
{
"text": "In order to refiect substitutions, two new Operations are introduced. The formerly forbidden case of nonterminal leafs now triggers the prediction of all possible substitution trees. On the other hand, the formerly end-test-only state of being at position \"ra\" for nonauxiliary roots now serves for the completion of predicted substitution trees.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "S-TAG Parser",
"sec_num": null
},
{
"text": "As modern workbenches (cf\" e.g., the workbench PAGE for Head-driven Phrase Structure Grammar [Netter, Oepen 97]) usually provide a generator, our parser is parametrised to work for generation according to the idea of bidirectional processing (cf\" e.g., [Neumann 941) .",
"cite_spans": [
{
"start": 253,
"end": 266,
"text": "[Neumann 941)",
"ref_id": null
}
],
"ref_spans": [],
"eq_spans": [],
"section": "S-TAG Generator",
"sec_num": null
},
{
"text": "As outlined by {Shieber et al. 90] a na'ive structuredriven top-down generator may not terminate (e.g. for genitive phrases in English and German). Furthermore the approach is inefficient because the input does not guide the gl!neration process. Instead of that, possible syntactic structures are realized and their corresponding logical forms are compared to the semantic input structure.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "S-TAG Generator",
"sec_num": null
},
{
"text": "A more natural way of guiding the generation process is to make it driven by the semantic input structure (indexing on meaning instead of indexing on string position). Generally speaking such generator predicts semantic heads. Two different procedures continue searching for a connection to sub-and the superderiviation tree.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "S-TAG Generator",
"sec_num": null
},
{
"text": "In the terrninology of [Shieber et al. 90] the generator predicts pivots. A pivot is defined as the lowest node in the tree such that it and all higher nodes up to the root node or a higher pivot node have the same semantics. According to the definition of a pivot node the set of grammar rules consists of two subsets. The set of chain rules consists of all rules in which the semantics of sorne right-hand side elernent is identical to the semantics of the left-hand side. The right-hand side element is called the semantic head. The set of non-chain rules contains all rules which do not satisfy 'this condition. The traversal will work top-down from the pivot node only using non-chain rules whereas the bottom-up steps which connect the pivot node with the root node only use chain rules.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "S-TAG Generator",
"sec_num": null
},
{
"text": "Adapting this mechanism to the generation of lexicalized TAGs means that the chain rules are corn\u2022 pletely deterrnined by the elementary tree under consideration 4 . Adjoining and substituLion rnpresent the application of non-chain rules. In order to illustrate this kind of processing let us assume that the input structure is (frequently{see(John,friends))). Furthermore, we assume that the grammar allows to pre-dict the trees described in Fig. 3 . Since bere is not the space to outline the specification lists of the individual nodes, the semantics of the trees is informally annotated at the nodes where x and y are variables to be filled during the unification at thut node. In a first step all predictible pivots according to the input structure can be written to the one and only item set during processing. This construction represents tbe unordered processing of the semantic structure. The bracketing structure of the logical form is achieved by evaluating the semantic expression associated with each elementary tree (e.g. for tree a 1 mod(x) wbere x is a value filled by the subtree of the foot node. The processing is successful only if a derivation tree can be constructed wbere all elements of the logical form occur only once 6 \u2022 Concerning the example two realizations for tbe input specification can be produced. The processing of the one with the sentential adverb (adjoing of a1) is obvious whereas the adjoining of a2 is not so clear. lt also works because the variable x at the foot node is unified with the VP node of h wltere according to the pivot definition the semantics on the spine from the root to the V node is identical. So, x contains the whole expression (see(John,friends)) and the check whether tbe bracketing structure is correct (i.e. the dependencies, specified in the logical form), is successful as weil. ",
"cite_spans": [
{
"start": 1244,
"end": 1247,
"text": "6 \u2022",
"ref_id": null
}
],
"ref_spans": [
{
"start": 443,
"end": 449,
"text": "Fig. 3",
"ref_id": "FIGREF2"
}
],
"eq_spans": [],
"section": "S-TAG Generator",
"sec_num": null
},
{
"text": "Here, one can decide whether the structures are collapsed, although their features may differ. In the fust case the disjunction of both feature descriptions is stored together with the history where they originally helonged to. Accordingly, more condensed structures are produced but the interpretation of the feature structures becomes more complicated.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
},
{
"text": "This transformation does not show the unification structures (c.f. footnote 1).",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
},
{
"text": "Due to the possibility of arbitrary mix-ups of precedences of children by REs, the expressions \"left/Tight to\" are to be understood in a more temporal tha.n local ma.nner, i.e. \"left of the foot node\" encloses all those items tha.t ha.ve been compute<l before computing the foot.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
},
{
"text": "Since empty semantic hea.ds can be associa.ted with their syntactic rea.lization they can be processecl in the same manner.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
},
{
"text": "Since the bracketing structure is tested explicitely during the combination of elementary trees the accepting condition can be weaker so that the logical form equivalence problem (cf. [Shieber 93]) does not occur here.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
}
],
"back_matter": [
{
"text": "(cf. [Harbusch 94)) where the size of structures influence the time in which the processing can be finished and results can be handed over to other components.Another topic of current considerations is how to define LD/LP-Schema-TAG which are especially interesting for gen.eration. We assume that it suffices to rewrite the NEXT function to adapt our parser to run LD/LP-Schema-TAGs on the structural level. Our suggestion is that the separation of structural combination and linear ordering saves processing time, especially for generation.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "acknowledgement",
"sec_num": null
}
],
"bib_entries": {
"BIBREF0": {
"ref_id": "b0",
"title": "XTAG System -A Wide Coverage Grammar for English",
"authors": [
{
"first": "C",
"middle": [],
"last": "Doran",
"suffix": ""
},
{
"first": "D",
"middle": [],
"last": "Egedi",
"suffix": ""
},
{
"first": "B",
"middle": [
"A"
],
"last": "Hockey",
"suffix": ""
},
{
"first": "B",
"middle": [],
"last": "Srinivas",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Zaidel",
"suffix": ""
},
{
"first": "; J",
"middle": [],
"last": "Gosling",
"suffix": ""
},
{
"first": "B",
"middle": [],
"last": "Joy",
"suffix": ""
},
{
"first": "G",
"middle": [],
"last": "Steele",
"suffix": ""
}
],
"year": 1994,
"venue": "the Proceedings of the 15th COLING",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "C. Doran, D. Egedi, B.A. Hockey, B. Srinivas, M. Zaidel. XTAG System -A Wide Coverage Grammar for English. In the Proceedings of the 15th COLING, Kyoto/Japan, 1994. [Gosling et al. 98] J. Gosling, B. Joy, and G. Steele. The Jaua Language Specification Addison-Wesley, Reading, 2nd. ed. 1998",
"links": null
},
"BIBREF1": {
"ref_id": "b1",
"title": "Incremental sentence processing with Schema-Tree Adjoining Grammars",
"authors": [
{
"first": "K",
"middle": [],
"last": "Harbusch",
"suffix": ""
}
],
"year": 1994,
"venue": "the procs. of the :Jrd International TAG+ Workshop",
"volume": "",
"issue": "",
"pages": "41--44",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "[Harbusch 94] K. Harbusch. Incremental sentence pro- cessing with Schema-Tree Adjoining Grammars. In the procs. of the :Jrd International TAG+ Work- shop, Paris, Frankreich, September 1994, TALANA- Report 94-01, pp. 41-44.",
"links": null
},
"BIBREF2": {
"ref_id": "b2",
"title": "The convergence of mildlycontextsensitive grammar formalisms",
"authors": [
{
"first": "A",
"middle": [
"K"
],
"last": "Joshi",
"suffix": ""
}
],
"year": 1986,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "A.K. Joshi. The convergence of mildly- contextsensitive grammar formalisms. In T. Wasow, P. Seils, eds., The Processing of Linguistic Structures. MIT-Press, Cambridge, MA/USA, 1986. [Netter, Oepen 97] K. Netter, S. Oepen. PAGE -Platform for Advanced Grammar Engineering.",
"links": null
},
"BIBREF3": {
"ref_id": "b3",
"title": "A Uniform Computational Model for Natural Langv.age Parsing and Generation",
"authors": [
{
"first": ";",
"middle": [
"G"
],
"last": "Sildes",
"suffix": ""
},
{
"first": "",
"middle": [
"M"
],
"last": "Neumann ; S",
"suffix": ""
},
{
"first": "F",
"middle": [
"C N"
],
"last": "Shieber",
"suffix": ""
},
{
"first": "G",
"middle": [],
"last": "Pereira",
"suffix": ""
},
{
"first": "R",
"middle": [
"C"
],
"last": "Van \u2022 Noord",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Moore",
"suffix": ""
}
],
"year": 1990,
"venue": "Shieber 93) S.M. Sltieber",
"volume": "16",
"issue": "",
"pages": "179--190",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Sildes (at http://cl-www.dfk.i.uni-sb.de/pagegifs/sli- des.html), Saarbr\u00fccken/Germany, 1997. [Neumann 94] G. Neumann. A Uniform Computa- tional Model for Natural Langv.age Parsing and Gen- eration. PhD thesis, Saarbr\u00fccken, Germany, 1994. [Shieber et al. 90] S.M. Shieber, F.C.N. Pereira, G. van \u2022 Noord, R.C. Moore. Semantic-Head-Driuen Genera- tion. Computational Linguistics, 16(1): 30-42, 1990. .[Shieber 93) S.M. Sltieber. The Problem of Logical- Form Equiuafonce. Computational Linguistics, 19(1): 179-190, 1993.",
"links": null
},
"BIBREF4": {
"ref_id": "b4",
"title": "Mathematicul and Computationai Aspects of Lexicalized Grammars",
"authors": [
{
"first": "Y",
"middle": [],
"last": "Schabes",
"suffix": ""
}
],
"year": 1990,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Y. Schabes. Mathematicul and Computa- tionai Aspects of Lexicalized Grammars. PhD thesis, Philadelphia, PA/USA, 1990.",
"links": null
},
"BIBREF5": {
"ref_id": "b5",
"title": "Characterising Mildly Context-Sensitive Grammar Formalisms",
"authors": [
{
"first": "D",
"middle": [],
"last": "Weir",
"suffix": ""
}
],
"year": 1987,
"venue": "J. Woch, F. Widmann, l{. Harbusch. A Reversible Approach to Parsing and Generation of Schema-TAGs. Teclmical Report",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "D. Weir. Characterising Mildly Context- Sensitive Grammar Formalisms. PhD. Proposal, Uni- versity of Pennsylvania, Philadelphia/USA, 1987. [Wach et al. 98] J. Woch, F. Widmann, l{. Harbusch. A Reversible Approach to Parsing and Generation of Schema-TAGs. Teclmical Report, University of Koblenz, forthcoming.",
"links": null
}
},
"ref_entries": {
"FIGREF0": {
"type_str": "figure",
"uris": null,
"text": "(WNP\" . \"\")) llJ.121+121+121.1a1 + lll-141 + 151 + J6l.[51) (({{\"DetP\" . '\"')) :substp T)) ((((\"N\" . \"\")) :headp T)) ((((''S\". \"\")) :substp T)) ((((\"N\u00b0\" . \"r'')) :suhstp T)) ({((\"G\" . \"\")) :headp T)) {(((\"NP\" . \"\")) :substp T))) ({{(\"N\u00b0\" . \"r\")) llf.[21 :constraints \"NA\" :constraint-type :NA) ((((\"N\" . \"\" )) :headp T)) {{WS\" . \"\")) :substp T))) Gramma.r transformation next step the annotation of the root node of sx is constructed by swnming up all alternatives according to b1, \"., b 0 where all labels are rewritten as numerical references pointing at the respective child.",
"num": null
},
"FIGREF2": {
"type_str": "figure",
"uris": null,
"text": "Predictible pivots",
"num": null
},
"TABREF0": {
"type_str": "table",
"num": null,
"content": "<table><tr><td>and its root node is annotated</td></tr><tr><td>with \"lll + IIl.121\" \u2022</td></tr><tr><td>Description of the S-TAG Workbench</td></tr><tr><td>In the following, the components of an S-TAG work-</td></tr><tr><td>bench (STAGWB) are outlined.</td></tr></table>",
"html": null,
"text": "In the first subsection a facility to transform arbitrary TAG grammars (in our case the UPENN tree bench [Daran et al. 94]) into schematic trees. Then the reversible component for parsing and generation is outlined (for details s. [Woch et al. 98])."
},
"TABREF1": {
"type_str": "table",
"num": null,
"content": "<table><tr><td>All</td><td>modules</td><td>are</td><td>implemented</td><td>in</td></tr><tr><td>JAVA [</td><td/><td/><td/><td/></tr></table>",
"html": null,
"text": "Gosling et al. 98]. Currently we run our transformation module to build a Schema-TAG equivalent test how the average runtime varies for TAGs and Schemu-TAGs. The differing size and depth of elementary trees is of special interest in incremental generation"
}
}
}
} |