File size: 33,502 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 |
{
"paper_id": "W12-0108",
"header": {
"generated_with": "S2ORC 1.0.0",
"date_generated": "2023-01-19T04:12:56.392858Z"
},
"title": "PRESEMT: Pattern Recognition-based Statistically Enhanced MT",
"authors": [
{
"first": "George",
"middle": [],
"last": "Tambouratzis",
"suffix": "",
"affiliation": {},
"email": ""
},
{
"first": "Marina",
"middle": [],
"last": "Vassiliou",
"suffix": "",
"affiliation": {},
"email": ""
},
{
"first": "Sokratis",
"middle": [],
"last": "Sofianopoulos",
"suffix": "",
"affiliation": {},
"email": ""
}
],
"year": "",
"venue": null,
"identifiers": {},
"abstract": "This document contains a brief presentation of the PRESEMT project that aims in the development of a novel language-independent methodology for the creation of a flexible and adaptable MT system.",
"pdf_parse": {
"paper_id": "W12-0108",
"_pdf_hash": "",
"abstract": [
{
"text": "This document contains a brief presentation of the PRESEMT project that aims in the development of a novel language-independent methodology for the creation of a flexible and adaptable MT system.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Abstract",
"sec_num": null
}
],
"body_text": [
{
"text": "The PRESEMT project constitutes a novel approach to the machine translation task. This approach is characterised by (a) introducing crossdisciplinary techniques, mainly borrowed from the machine learning and computational intelligence domains, in the MT paradigm and (b) using relatively inexpensive language resources. The aim is to develop a language-independent methodology for the creation of a flexible and adaptable MT system, the features of which ensure easy portability to new language pairs or adaptability to particular user requirements and to specialised domains with minimal effort. PRESEMT falls within the Corpus-based MT (CBMT) paradigm, using a small bilingual parallel corpus and a large TL monolingual corpus. Both these resources are collected as far as possible over the web, to simplify the development of resources for new language pairs.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1."
},
{
"text": "The main aim of PRESEMT has been to alleviate the reliance on specialised resources. In comparison, Statistical MT requires large parallel corpora for the source and target languages. PRESEMT relaxes this requirement by using a small parallel corpus, augmented by a large TL monolingual corpus.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1."
},
{
"text": "The PRESEMT system is distinguished into three stages, as shown in Figure 1: 1. Pre-processing stage: This is the stage where the essential resources for the MT system are compiled. It consists of four discrete modules: (a) the Corpus creation & annotation module, being responsible for the compilation of monolingual and bilingual corpora over the web and their annotation; (b) the Phrase aligner module, which processes a bilingual corpus to perform phrasal level alignment within a language pair; (c) the Phrasing model generator that elicits an SL phrasing model on the basis of the aforementioned alignment and employs it as a parsing tool during the translation process; (d) the Corpus modelling module, which creates semanticsbased TL models used for disambiguation purposes during the translation process.",
"cite_spans": [],
"ref_spans": [
{
"start": 67,
"end": 76,
"text": "Figure 1:",
"ref_id": null
}
],
"eq_spans": [],
"section": "PRESEMT system structure",
"sec_num": "2."
},
{
"text": "The translation in PRESEMT is a top-down two-phase process, distinguished into the Structure selection module, where the constituent phrases of an SL sentence are reordered according to the TL, and the Translation equivalent selection module where translation disambiguation is resolved and word order within phrases is established. Closely integrated to the translation engine, but not part of the main translation process, is the Optimisation module, which is responsible for automatically improving the performance of the two translation phases by fine-tuning the values of the various system parameters.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Main translation engine:",
"sec_num": "2."
},
{
"text": "The third stage is useroriented and comprises (i) the Post-processing and (ii) the User Adaptation modules. The first module allows the user to modify the systemgenerated translations towards their requirements. The second module enables PRESEMT to adapt to this input so that it learns to generate translations closer to the users' requirements. The postprocessing stage represents work in progress to be reported in future publications, the present article focussing on the actual strategy for generating the translation.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Post-processing stage:",
"sec_num": "3."
},
{
"text": "The bilingual corpus contains literal translations, to allow the extrapolation of mapping information from SL to TL, though this may affect the translation quality. The Phrase aligner module (PAM) performs offline SL -TL word and phrase alignment within this corpus. PAM serves as a language-independent method for mapping corresponding terms within a language pair, by circumventing the problem of achieving compatibility between the outputs of two different parsers, one for the SL and one for the TL. PAM relies on a single parser for the one language and generates an appropriate phrasing model for the other language in an automated manner. The phrases are assumed to be flat and linguistically valid. As a parser, any available tool may be used (the TreeTagger (Schmid, 1994) is used in the present implementation for English). PAM processes a bilingual corpus of SL -TL sentence pairs, taking into account the parsing information in one language (in the current implementation the TL side) and making use of a bilingual lexicon and information on potential phrase heads; the output being the bilingual corpus aligned at word, phrase and clause level. Thus, at a phrasal level, the PAM output indicates how an SL structure is transformed into the TL. For instance, based on a sentence pair from the parallel corpus, the SL sentence with structure A-B-C-D is transformed into A'-C'-D'-B', where X is a phrase in SL and X' is a phrase in TL. Further PAM details are reported in Tambouratzis et al. (2011).",
"cite_spans": [
{
"start": 767,
"end": 781,
"text": "(Schmid, 1994)",
"ref_id": "BIBREF5"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Processing of the bilingual corpus",
"sec_num": "3."
},
{
"text": "The PAM output in terms of SL phrases is then handed over to the Phrasing model generator (PMG), which is trained to determine the phrasal structure of an input sentence. PMG reads the SL phrasing as defined by PAM and generates an SL phrasing model using a probabilistic methodology. This phrasing model is then applied in segmenting any arbitrary SL text being input to the PRESEMT system for translation. PMG is based on the Conditional Random Fields model (Lafferty et al., 1999) which has been found to provide the highest accuracy. The SL text segmented into phrases by PMG is then input to the 1 st translation phase. For a new language pair, the PAM-PMG chain is implemented without any manual correction of outputs.",
"cite_spans": [
{
"start": 460,
"end": 483,
"text": "(Lafferty et al., 1999)",
"ref_id": null
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Processing of the bilingual corpus",
"sec_num": "3."
},
{
"text": "The language models created by the Corpus modelling module can only serve translation dis-ambiguation purposes; thus another form of interfacing with the monolingual corpus is essential for the word reordering task within each phrase. The size of the data accessed is very large. Typically, a monolingual corpus contains 3 billion words, 10 8 sentences and approximately 10 9 phrases. Since the models for the TL phrases need to be accessed in real-time to allow word reordering within each phrase, the module uses the phrase indexed representation of the monolingual corpus. This phrase index is created based on four criteria: (i) phrase type, (ii) phrase head lemma, (iii) phrase head PoS tag and (iv) number of tokens in the phrase.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Organising the monolingual corpus",
"sec_num": "4."
},
{
"text": "Indexing is performed by extracting all phrases from the monolingual corpus, each of which is transformed to the java object instance used within the PRESEMT system. The phrases are then organised in a hash map that allows multiple values for each key, using as a key the 4 aforementioned criteria. Statistical information about the number of occurrences of each phrase in the corpus is also included. Finally, each map is serialised and stored in the appropriate file in the PRESEMT path, with each file being given a suitable name for easy retrieval. For example, for the English monolingual corpus, all verb phrases with head lemma \"read\" (verb) and PoS tag \"VV\" containing 2 tokens in total are stored in the file \"Corpora\\EN\\Phrases\\VC\\read_VV\". If any of these criteria has a different value, then a separate file is created (for instance for verb phrases with head \"read\" that contain 3 tokens).",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Organising the monolingual corpus",
"sec_num": "4."
},
{
"text": "The PRESEMT translation process entails first the establishment of the sentence phrasal structure and then the resolution of the intra-phrasal arrangements, i.e. specifying the correct word order and deciding upon the appropriate candidate translation. Both phases involve searching for suitable matching patterns at two different levels of granularity, the first (coarse-grained) aiming at defining a TL-compatible ordering of phrases in the sentence and the second (finegrained) determining the internal structure of phrases. While the first phase utilises the small bilingual corpus, the second phase makes use of the large monolingual corpus. To reduce the translation time required, both corpora are processed in advance and the processed resources are stored in such a form as be retrieved as rapidly as possible during translation.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Main translation engine",
"sec_num": "5."
},
{
"text": "Each SL sentence input for translation is tagged and lemmatised and then it is segmented into phrases by the Phrasing model generator on the basis of the SL phrasing model previously created. For establishing the correct phrase order according to the TL, the parallel corpus needs to be pre-processed using the Phrase aligner module to identify word and phrase alignments between the equivalent SL and TL sentences. During structure selection, the SL sentence is aligned to each SL sentence of the parallel corpus, as processed by the PAM and assigned a similarity score using an algorithm from the dynamic programming paradigm. The similarity score is calculated by taking into account edit operations (replacement, insertion or removal) needed to be performed in the input sentence in order to transform it to the corpus SL sentence. Each of these operations has an associated cost, considered as a system parameter. The aligned corpus sentence that achieves the highest similarity score is the most similar one to the input source sentence. This comparison process relies on a set of similarity parameters (e.g. phrase type, phrase head etc.), the values of which are optimised by employing the optimisation module.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Translation Phase 1: Structure selection module",
"sec_num": "5.1"
},
{
"text": "The implementation is based on the Smith-Waterman algorithm (Smith and Waterman, 1981) , initially proposed for determining similar regions between two protein or DNA sequences. The algorithm is guaranteed to find the optimal local alignment between the two input sequences at clause level.",
"cite_spans": [
{
"start": 60,
"end": 86,
"text": "(Smith and Waterman, 1981)",
"ref_id": "BIBREF6"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Translation Phase 1: Structure selection module",
"sec_num": "5.1"
},
{
"text": "After establishing the order of phrases within each sentence, the second phase of the translation process is initiated, comprising two distinct tasks. The first task is to resolve the lexical ambiguity, by picking one lemma from each set of possible translations (as provided by a bilingual dictionary). In doing so, this module makes use of the semantic similarities between words which have been determined by the Corpus Modelling module through a co-occurrence analysis on the monolingual TL corpus. That way, the best combination of lemmas from the sets of candidate translations is determined for a given context. In the second task, the most similar phrases to the TL structure phrases are retrieved from the monolingual corpus to provide local structural information such as word-reordering. A matching algorithm selects the most similar from the set of the retrieved TL phrases through a comparison process, which is viewed as an assignment problem, using the Gale-Shapley algorithm (Gale and Shapley, 1962) .",
"cite_spans": [
{
"start": 991,
"end": 1015,
"text": "(Gale and Shapley, 1962)",
"ref_id": "BIBREF0"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Translation Phase 2: Translation equivalent selection module",
"sec_num": "5.2"
},
{
"text": "To date MT systems based on the PRESEMT methodology have been created for a total of 8 languages, indicating the flexibility of the proposed approach. Table 1 illustrates an indicative set of results obtained by running automatic evaluation metrics on test data translated by the 1 st PRESEMT prototype for a selection of language pairs, due to space restrictions.",
"cite_spans": [],
"ref_spans": [
{
"start": 151,
"end": 158,
"text": "Table 1",
"ref_id": null
}
],
"eq_spans": [],
"section": "Experiments & evaluation results",
"sec_num": "6."
},
{
"text": "In the case of the language pair English-to-German, these results are contrasted to the ones obtained when translating the same test set with Moses (Koehn et al., 2007) .It is observed that for the English-to-German language pair, PRESEMT achieved approximately 50% of the MOSES BLEU score and 80% of the MOSES with respect to the Meteor and TER scores. These are reasonably competitive results compared to an established system such as Moses. Furthermore, it should taken into consideration that (a) the PRESEMT results were obtained by the 1 st system prototype, (b) PRESEMT is still under development and (c) only one reference translation was used per sentence.",
"cite_spans": [
{
"start": 148,
"end": 168,
"text": "(Koehn et al., 2007)",
"ref_id": "BIBREF1"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Experiments & evaluation results",
"sec_num": "6."
},
{
"text": "Newer versions of the PRESEMT system, incorporating more advanced versions of the different modules are expected to result in substantially improved translation accuracies. In particular, the second translation phase will be further researched. In addition, experiments have indicated that the language modelling module can provide additional improvement in the performance. Finally, refinements in PAM and PMG may lead in increased translation accuracies.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Experiments & evaluation results",
"sec_num": "6."
},
{
"text": "Find out more about the project on the PRE-SEMT website: www.presemt.eu. Also, the PRESEMT prototype may be tried at: ",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Links",
"sec_num": "7."
}
],
"back_matter": [
{
"text": "The research leading to these results has received funding from the European Community's Seventh Framework Programme (FP7/2007-2013 under grant agreement n\u00b0 248307.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Acknowledgments",
"sec_num": null
}
],
"bib_entries": {
"BIBREF0": {
"ref_id": "b0",
"title": "College Admissions and the Stability of Marriage",
"authors": [
{
"first": "D",
"middle": [],
"last": "Gale",
"suffix": ""
},
{
"first": "L",
"middle": [
"S"
],
"last": "Shapley",
"suffix": ""
}
],
"year": 1962,
"venue": "American Mathematical Monthly",
"volume": "69",
"issue": "",
"pages": "9--14",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Gale D. and L. S. Shapley. 1962. College Admissions and the Stability of Marriage. American Mathe- matical Monthly, Vol. 69, pp. 9-14.",
"links": null
},
"BIBREF1": {
"ref_id": "b1",
"title": "Moses: Open Source Toolkit for Statistical Machine Translation",
"authors": [
{
"first": "P",
"middle": [],
"last": "Koehn",
"suffix": ""
},
{
"first": "H",
"middle": [],
"last": "Hoang",
"suffix": ""
},
{
"first": "A",
"middle": [],
"last": "Birch",
"suffix": ""
},
{
"first": "C",
"middle": [],
"last": "Callison-Burch",
"suffix": ""
},
{
"first": "M",
"middle": [],
"last": "Federico",
"suffix": ""
},
{
"first": "N",
"middle": [],
"last": "Bertoldi",
"suffix": ""
},
{
"first": "B",
"middle": [],
"last": "Cowan",
"suffix": ""
},
{
"first": "W",
"middle": [],
"last": "Shen",
"suffix": ""
},
{
"first": "C",
"middle": [],
"last": "Moran",
"suffix": ""
},
{
"first": "R",
"middle": [],
"last": "Zens",
"suffix": ""
},
{
"first": "C",
"middle": [],
"last": "Dyer",
"suffix": ""
},
{
"first": "O",
"middle": [],
"last": "Bojar",
"suffix": ""
},
{
"first": "A",
"middle": [],
"last": "Constantin",
"suffix": ""
},
{
"first": "E",
"middle": [],
"last": "Herbst",
"suffix": ""
}
],
"year": 2007,
"venue": "Proceedings of the ACL-2007 Demo and Poster Sessions",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Koehn P., H. Hoang, A. Birch, C. Callison-Burch, M. Federico, N. Bertoldi, B. Cowan, W. Shen, C. Moran, R. Zens, C. Dyer, O. Bojar, A. Constantin, and E. Herbst. 2007. Moses: Open Source Toolkit for Statistical Machine Translation. In Proceedings of the ACL-2007 Demo and Poster Sessions.",
"links": null
},
"BIBREF2": {
"ref_id": "b2",
"title": "The Hungarian method for the assignment problem",
"authors": [
{
"first": "H",
"middle": [
"W"
],
"last": "Kuhn",
"suffix": ""
}
],
"year": 1955,
"venue": "Naval Research Logistics Quarterly",
"volume": "2",
"issue": "",
"pages": "83--97",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Kuhn H. W. 1955. The Hungarian method for the as- signment problem. Naval Research Logistics Quar- terly, Vol. 2, pp.83-97.",
"links": null
},
"BIBREF3": {
"ref_id": "b3",
"title": "Conditional Random Fields: Probabilistic Models for Segmenting and Labelling Sequence Data. Proceedings of ICML Conference",
"authors": [
{
"first": "J",
"middle": [],
"last": "Lafferty",
"suffix": ""
},
{
"first": "A",
"middle": [],
"last": "Mccallum",
"suffix": ""
},
{
"first": "F",
"middle": [],
"last": "Pereira",
"suffix": ""
}
],
"year": 2001,
"venue": "",
"volume": "",
"issue": "",
"pages": "282--289",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Lafferty J., A. McCallum, F. Pereira. 2001. Condi- tional Random Fields: Probabilistic Models for Segmenting and Labelling Sequence Data. Pro- ceedings of ICML Conference, pp.282-289.",
"links": null
},
"BIBREF4": {
"ref_id": "b4",
"title": "Algorithms for the assignment and transportation problems",
"authors": [
{
"first": "J",
"middle": [],
"last": "Munkres",
"suffix": ""
}
],
"year": 1957,
"venue": "Journal of the Society for Industrial and Applied Mathematics",
"volume": "5",
"issue": "",
"pages": "32--38",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Munkres J. 1957. Algorithms for the assignment and transportation problems. Journal of the Society for Industrial and Applied Mathematics, Vol. 5, pp.32- 38.",
"links": null
},
"BIBREF5": {
"ref_id": "b5",
"title": "Probabilistic Part-of-Speech Tagging Using Decision Trees",
"authors": [
{
"first": "H",
"middle": [],
"last": "Schmid",
"suffix": ""
}
],
"year": 1994,
"venue": "Proceedings of International Conference on New Methods in Language Processing",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Schmid, H. 1994. Probabilistic Part-of-Speech Tag- ging Using Decision Trees, Proceedings of Interna- tional Conference on New Methods in Language Processing, Manchester, UK.",
"links": null
},
"BIBREF6": {
"ref_id": "b6",
"title": "Identification of Common Molecular Subsequences",
"authors": [
{
"first": "T",
"middle": [
"F"
],
"last": "Smith",
"suffix": ""
},
{
"first": "M",
"middle": [
"S"
],
"last": "Waterman",
"suffix": ""
}
],
"year": 1981,
"venue": "Journal of Molecular Biology",
"volume": "147",
"issue": "",
"pages": "195--197",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Smith T. F. and M. S. Waterman. 1981. Identification of Common Molecular Subsequences. Journal of Molecular Biology, 147: 195-197.",
"links": null
}
},
"ref_entries": {
"TABREF0": {
"type_str": "table",
"html": null,
"text": "presemt.cslab.ece.ntua.gr:8080/presemt_interface_test PRESEMT Evaluation results for different language pairs.",
"num": null,
"content": "<table><tr><td/><td/><td/><td/><td colspan=\"4\">Tambouratzis G., F. Simistira, S. Sofianopoulos, N.</td></tr><tr><td/><td/><td/><td/><td colspan=\"4\">Tsimboukakis and M. Vassiliou 2011. A resource-</td></tr><tr><td/><td/><td/><td/><td colspan=\"4\">light phrase scheme for language-portable MT,</td></tr><tr><td/><td/><td/><td/><td colspan=\"4\">Proceedings of the 15 th International Conference of</td></tr><tr><td/><td/><td/><td/><td colspan=\"4\">the European Association for Machine Translation,</td></tr><tr><td/><td/><td/><td/><td colspan=\"4\">30-31 May 2011, Leuven, Belgium, pp. 185-192.</td></tr><tr><td colspan=\"2\">Table 1 -Language Pair</td><td colspan=\"2\">Sentence set</td><td/><td colspan=\"2\">Metrics</td></tr><tr><td>SL</td><td>TL</td><td colspan=\"2\">Number Source</td><td>BLEU</td><td>NIST</td><td>Meteor</td><td>TER</td></tr><tr><td>English</td><td>German</td><td>189</td><td>web</td><td colspan=\"2\">0.1052 3.8433</td><td>0.1939</td><td>83.233</td></tr><tr><td>German</td><td>English</td><td>195</td><td>web</td><td colspan=\"2\">0.1305 4.5401</td><td>0.2058</td><td>74.804</td></tr><tr><td>Greek</td><td>English</td><td>200</td><td>web</td><td colspan=\"2\">0.1011 4.5124</td><td>0.2442</td><td>79.750</td></tr><tr><td>English</td><td>German</td><td>189</td><td>web</td><td colspan=\"2\">0.2108 5.6517</td><td>0.2497</td><td>68.190 Moses</td></tr><tr><td/><td/><td colspan=\"4\">Figure 1 -PRESEMT system architecture.</td><td/></tr><tr><td/><td/><td/><td>WEB</td><td/><td/><td/></tr><tr><td/><td/><td>Offline corpora</td><td colspan=\"2\">Corpus creation</td><td/><td/></tr><tr><td/><td/><td>creation</td><td colspan=\"2\">& annotation</td><td/><td/></tr><tr><td/><td/><td/><td>module</td><td/><td/><td/></tr><tr><td/><td/><td/><td/><td/><td colspan=\"2\">TL Monolingual</td></tr><tr><td/><td/><td/><td/><td/><td colspan=\"2\">Corpus &</td></tr><tr><td/><td/><td/><td/><td/><td colspan=\"2\">Corpus model</td></tr><tr><td/><td/><td/><td colspan=\"2\">Corpus modelling</td><td/><td/></tr><tr><td/><td/><td/><td>module</td><td/><td/><td/></tr><tr><td/><td/><td/><td/><td/><td colspan=\"2\">Bilingual parallel</td></tr><tr><td/><td/><td/><td colspan=\"2\">Phrase aligner</td><td colspan=\"2\">corpus</td></tr><tr><td/><td>Bilingual</td><td/><td>module</td><td/><td/><td/><td>Post-processing</td></tr><tr><td/><td>Lexicon</td><td/><td/><td/><td/><td/><td>& User</td></tr><tr><td/><td/><td/><td colspan=\"2\">Phrasing model generator</td><td colspan=\"2\">Bilingual aligned corpus</td><td>adaptation modules</td></tr><tr><td/><td/><td colspan=\"3\">Phase 1: Structure selection</td><td colspan=\"2\">Phase 2: Translation equivalent</td></tr><tr><td>S SL L t te ex xt t</td><td>Tagging-Lemmatising, Lexicon look-up &</td><td colspan=\"3\">Phrase reordering of each sentence in the SL text</td><td colspan=\"2\">selection Word reordering & disambiguation</td><td>F Fi in na al l tr ra an ns sl la at ti io on n t</td></tr><tr><td/><td>Token generation</td><td/><td/><td/><td/><td/></tr><tr><td/><td/><td/><td colspan=\"3\">Optimisation module</td><td/></tr></table>"
}
}
}
} |