| { | |
| "paper_id": "W98-0101", | |
| "header": { | |
| "generated_with": "S2ORC 1.0.0", | |
| "date_generated": "2023-01-19T06:03:50.960193Z" | |
| }, | |
| "title": "An Experiment on Synchronous TAGs for the Construction of a Transfer Module", | |
| "authors": [ | |
| { | |
| "first": "Alexandre", | |
| "middle": [], | |
| "last": "Agustini", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "Vera Lucia Strube de Lima PUCRS -Instituto de Inform\u00e4tica Av. Ipiranga", | |
| "location": { | |
| "addrLine": "6681 predio 30 bloco 4", | |
| "postCode": "90619-900", | |
| "settlement": "Porto Alegre RS -BRASIL" | |
| } | |
| }, | |
| "email": "[email protected]" | |
| } | |
| ], | |
| "year": "", | |
| "venue": null, | |
| "identifiers": {}, | |
| "abstract": "This paper presents some considerations on the use of Synchronous T AGs for the design of a structural transfer module, which is the main component of transfer-based systems for Machine Translation. The transfer module establishes the correspondences between the structural representation of both the source and target languages. A study of a corpus from Economics was carried out in order to define structural divergences for the translation between the Portuguese and English languages.", | |
| "pdf_parse": { | |
| "paper_id": "W98-0101", | |
| "_pdf_hash": "", | |
| "abstract": [ | |
| { | |
| "text": "This paper presents some considerations on the use of Synchronous T AGs for the design of a structural transfer module, which is the main component of transfer-based systems for Machine Translation. The transfer module establishes the correspondences between the structural representation of both the source and target languages. A study of a corpus from Economics was carried out in order to define structural divergences for the translation between the Portuguese and English languages.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Abstract", | |
| "sec_num": null | |
| } | |
| ], | |
| "body_text": [ | |
| { | |
| "text": "Machine translation (MT) has been a challenge for linguists and computer scientists over the last decades. During this period, plenty of progress was accomplished, though the results are not yet the ones expected.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Introduction", | |
| "sec_num": "1" | |
| }, | |
| { | |
| "text": "Transfer based approaches to MT involve three main phases: analysis, . transfer and generation. During analysis, the syntactic and semantic' structure of a sentence is made explicit through a source language (SL) grammar and semantic processing modules. The result of the analysis is one or more syntactic and semantic representations which are used to construct a syntactic and/or semantic representation in the target language (TL) through a series of transfer rules and according to a bilingual lexicon. ", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Introduction", | |
| "sec_num": "1" | |
| }, | |
| { | |
| "text": "Our basic approach is corpus-driven. We started by collecting a source-language corpus (Portuguese sentences) in a limited domain. The corpus made up by 200 sentences was created randomly from an economics headlines database. About 5 0% of them were discarded because they were ill-formed senteces. The database had previously been generated from a news broadcasting system.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Corpus based development", | |
| "sec_num": "2" | |
| }, | |
| { | |
| "text": "An English version of the corpus was produced by a native translator \u2022 with experience in the domain tenninology. Finally, both corpuses were tagged and aligned in order to achieve:", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Corpus based development", | |
| "sec_num": "2" | |
| }, | |
| { | |
| "text": "\u2022 virtual grammars 1 for both the source and target corpuses: the subset of lexicalized trees necessary for syntactic/semantic analysis of source and target corpus was defincd. These grammars are based on [ \u2022 translation discrepancies: translation problems to be solved during transfer from source to target structures were addressed.", | |
| "cite_spans": [ | |
| { | |
| "start": 205, | |
| "end": 206, | |
| "text": "[", | |
| "ref_id": null | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Corpus based development", | |
| "sec_num": "2" | |
| }, | |
| { | |
| "text": "We found it helpful to divide translation problems into three different types: lexical, syntactic and lexical-semantic. These terms are used according to the following concept (according to [Dorr 94] ): Lexical problems are concemed about finding correct choices for expressions that occur in the source and target languages. Syntactic problems feature syntactic properties associated with each language (i.e\" properties that are independent of the actual lexical items that are used).", | |
| "cite_spans": [ | |
| { | |
| "start": 190, | |
| "end": 199, | |
| "text": "[Dorr 94]", | |
| "ref_id": null | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Corpus based development", | |
| "sec_num": "2" | |
| }, | |
| { | |
| "text": "Finally, lexical-semantic problems which feature properties that are lexically determined.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Corpus based development", | |
| "sec_num": "2" | |
| }, | |
| { | |
| "text": "Same exarnples of divergences observed in the corpus are presented on Table 1 . In case ( I ), problems originated from lexical gaps in the source and target languages are shown; the translation has to deal with structural problems and feature inheritance. Syntactic problems (2) usually have to do with word ", | |
| "cite_spans": [], | |
| "ref_spans": [ | |
| { | |
| "start": 70, | |
| "end": 77, | |
| "text": "Table 1", | |
| "ref_id": "TABREF2" | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "Corpus based development", | |
| "sec_num": "2" | |
| }, | |
| { | |
| "text": "Figure l illustrates the proposed model in which the structural divergence related information is modeled in two different dictionaries: a structural dictionary and a bilingual dictionary. The structural dictionary connects L TAGs structures that define both the source and the target languages. This structure maintains the node-to-node links between the source and the target grammars. All elementary trees in the source language are associated with trees in the target language. lt has inforrnation about the inheritance of semantic attributes and also holds all the information for syntactic divergence resolutiori Table 2 illustrates some entries of the structural dictionary. The bilingual dictionary contains the rules for the resolution of lexical and lexical-semantic divergences. This dictionary manipulates the pairs of lexicalized items and points out one or more elementary structures of the structurai dictionary to which the item is anchor. In this dictionary, derivation tree fragments can be defined, with the purpose of resolvihg lexical and lexical-semantic divergences. Furthermore, the dictionary can extend the rules contained in the structural dictionary to state the restnctlons imposed by the accomplished lexical insertion. A fragment of the bilingual dictionary is presented in Table 3 .", | |
| "cite_spans": [], | |
| "ref_spans": [ | |
| { | |
| "start": 619, | |
| "end": 626, | |
| "text": "Table 2", | |
| "ref_id": "TABREF3" | |
| }, | |
| { | |
| "start": 1305, | |
| "end": 1313, | |
| "text": "Table 3", | |
| "ref_id": "TABREF5" | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "A Model Proposed and Implementation", | |
| "sec_num": "3" | |
| }, | |
| { | |
| "text": "The transfer module receives a sequence of lexical items, generated by a lexiconmorphologic module. The output corresponds to one or more derivation trees in the target language with all structural modifications accomplished and decorated with the semantic features inherited from the source language.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "A Model Proposed and Implementation", | |
| "sec_num": "3" | |
| }, | |
| { | |
| "text": "Virtual grammars for source and target languages are descrihed in an independent way and the notation introduced by [ Due to the incremental characteristic of the STAGs method, transfer functions were incorporated to syntactic analysis. The implementation involves two distinct steps: syntactic analysis (parser) and verification.", | |
| "cite_spans": [ | |
| { | |
| "start": 116, | |
| "end": 117, | |
| "text": "[", | |
| "ref_id": null | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "A Model Proposed and Implementation", | |
| "sec_num": "3" | |
| }, | |
| { | |
| "text": "The parser uses a top-down algorithm for L TAG recognition. Each operation carried out by the parser in the SL enables one or more operations in the TL. The output is: for each SL syntactic structure a set of structures in the TL is generated.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "A Model Proposed and Implementation", | |
| "sec_num": "3" | |
| }, | |
| { | |
| "text": "During the process of analysis and translation, two types of attributes are manipulated: structural and semantic attributes. The structural attributes are inherent to each language and do not need to be transferred. On the other hand, semantic attributes are inherited by each one of the accomplished items of the pairs in lexicalized trees.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "A Model Proposed and Implementation", | |
| "sec_num": "3" | |
| }, | |
| { | |
| "text": "Finally, in the verification step, the unification of semantic features and the verification for structural consistency on the generated target trees is carried out. This process is based on the target LT AG grammar and inconsistent trees are discarded. This work investigated the use of the STAGs fonnalism for the treatment of lexical, syntactic and lexical-semantic divergences defined from a corpus in the field of Economics. Due to the extended domain of locality of LTAGs, it is possible to define regular correspondences among complex structures without the need of intermediary representations.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "A Model Proposed and Implementation", | |
| "sec_num": "3" | |
| }, | |
| { | |
| "text": "Although it was possible to set the translation mies for about 85% of the selected corpus ( composed of 90 sentences ), the model cannot yet be validated due to the short number of sample sentences.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "A Model Proposed and Implementation", | |
| "sec_num": "3" | |
| }, | |
| { | |
| "text": "Nowadays, we are starting to work on tagging and aligning tools for a bilingual corpus. These tools will allow us to set a more complex corpus of sentences to validate the work we have developed.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "A Model Proposed and Implementation", | |
| "sec_num": "3" | |
| }, | |
| { | |
| "text": "In this context, virtual grammar refers to a syntactic structure subset necessary for parsing any input sentence and generating target structures occuring in the corpus.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "", | |
| "sec_num": null | |
| } | |
| ], | |
| "back_matter": [], | |
| "bib_entries": { | |
| "BIBREF0": { | |
| "ref_id": "b0", | |
| "title": "Experiencia de Utiliza9iio da Forma/ismo STAGs para a Constru9iio de um Modula de Transferencia Estrutural", | |
| "authors": [ | |
| { | |
| "first": "A", | |
| "middle": [], | |
| "last": "Agustlni", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "A", | |
| "middle": [], | |
| "last": "Abeille", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1990, | |
| "venue": "Proceedings of the 13 1 h International Conference an Computational Linguistics", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "AGUSTlNI, A. Experiencia de Utiliza9iio da Forma/ismo STAGs para a Constru9iio de um Modula de Transferencia Estrutural. Master's Degree in Informatics, PUCRS, RS, Brazil, Mater's Dissertation, 1997. [Abeille 90] ABEILLE, A. et. al. Using Lexicalized T AGs for Machine Translation. In Proceedings of the 13 1 h International Conference an Computational Linguistics, Helsinki, Finland, 1990. [Beckeret et al. 94] BECKER, T. et al. A Lexicalized TAG for English. University of Pennsylvania, Technical Report, i994.", | |
| "links": null | |
| }, | |
| "BIBREF1": { | |
| "ref_id": "b1", | |
| "title": "Machine Translation Divergences: A Formal Description and Proposed Solution. Computational Linguistics, v. 20, n. 4", | |
| "authors": [ | |
| { | |
| "first": "B", | |
| "middle": [ | |
| "J" | |
| ], | |
| "last": "Dorr", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "W", | |
| "middle": [ | |
| "J" | |
| ], | |
| "last": "Hutchins", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "H", | |
| "middle": [ | |
| "L" | |
| ], | |
| "last": "Somers", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1994, | |
| "venue": "", | |
| "volume": "1", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "DORR, B. J. Machine Translation Divergences: A Formal Description and Proposed Solution. Computational Linguistics, v. 20, n. 4, 1994. [Hutchins & Somers 92] HUTCHINS, W. J. & SOMERS, H. L. An lntroduction to Machine Translation. Academic Press, Great Britain, 1 992.", | |
| "links": null | |
| }, | |
| "BIBREF2": { | |
| "ref_id": "b2", | |
| "title": "Uma Experiencia de Utilizaqiio da Formalismo de Gramdticas de Aqjunqiio de Arvores para a Lingua Portuguesa", | |
| "authors": [ | |
| { | |
| "first": "K", | |
| "middle": [], | |
| "last": "Kipper", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1994, | |
| "venue": "CPGCC-UFRGS", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "[Kipper 94] KIPPER, K. Uma Experiencia de Utilizaqiio da Formalismo de Gramdticas de Aqjunqiio de Arvores para a Lingua Portuguesa. Master's Degree in lnformatics, CPGCC- UFRGS, Porto Alegre, RS, Brazil, Mater's Dissertation, 1994. [Rambow & Satta 96] RAMBOW, 0.", | |
| "links": null | |
| }, | |
| "BIBREF3": { | |
| "ref_id": "b3", | |
| "title": "Synchronous Tree-Adjoinin~ Gramrnars", | |
| "authors": [ | |
| { | |
| "first": "G", | |
| "middle": [], | |
| "last": "Satta", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "S", | |
| "middle": [ | |
| "M" | |
| ], | |
| "last": "Shieber", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Y", | |
| "middle": [], | |
| "last": "Shabes", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1990, | |
| "venue": "Proceedings of the 19 1 '1 International Conference an Computational Linguistics", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "SATTA, G. Synchronous Models of Language. In Proceedings of the 19 1 '1 International Conference an Computational Linguistics, California, USA, 1996. [Shieber & Shabes 90] SHIEBER, S. M. & SHABES, Y. Synchronous Tree- Adjoinin~ Gramrnars. In Proceedings of the 13 1 International Conference an Computational Linguistics, Helsinki, Finland, 1990.", | |
| "links": null | |
| }, | |
| "BIBREF4": { | |
| "ref_id": "b4", | |
| "title": "Bi-Lexical Rules for Multi-Lexeme Translation in Lexicalist MT. MI-95", | |
| "authors": [ | |
| { | |
| "first": "A", | |
| "middle": [], | |
| "last": "Trujillo", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1995, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "TRUJILLO, A. Bi-Lexical Rules for Multi-Lexeme Translation in Lexicalist MT. MI-95, Leuven, Belgium, 1995.", | |
| "links": null | |
| } | |
| }, | |
| "ref_entries": { | |
| "FIGREF0": { | |
| "uris": null, | |
| "num": null, | |
| "type_str": "figure", | |
| "text": "~~de\":-+~a. eX,.,\"\"\"'\"'10s ..,~;o ..... t;,,nf \"'t.. ... as0 the last one, a lexical-semantic problem, see (3), the Portuguese verb ifazer) and its complement (leil\u00e4o) are translated into a verb (to auction) in English. peso Mexicano Mexican peso bolsa de Nyork New York stock exchange(3) fazer leiliio to auction" | |
| }, | |
| "FIGREF1": { | |
| "uris": null, | |
| "num": null, | |
| "type_str": "figure", | |
| "text": "Overview of Proposed Model" | |
| }, | |
| "FIGREF2": { | |
| "uris": null, | |
| "num": null, | |
| "type_str": "figure", | |
| "text": "# format: # ( source_id : target_id ) -7 [ links ] # transitive verbs (NP object complement) # S( N V N) --> S( NP CV( V NP)) ( 2:302) = { $0:$0 ' $1 :$1 ' $2:$3 ' $3:$4 1 ; # adjectives # N( N Adj) --> NP( Adj NP ) (100:400) = { $0:$0' $1 :$2 '$2:$1 J; # N( Adj N ) --> NP( Adj NP) ~101 : 400) = { $0:$0' $1:$1 '$2:$2); # adjectival phrase .# +NPROP = Proper Noun # N( N Prep N ) --> NP( Adj NP ) ( 20:400) = [ $0:$0, $1:$2, $3[+NPROP]:$1];" | |
| }, | |
| "FIGREF3": { | |
| "uris": null, | |
| "num": null, | |
| "type_str": "figure", | |
| "text": "(fazer : ##) II redefinition of verb fazer % S(N V(#lex=fazer] N(N{#lex=leilao] Prep N)) (NP CV ( V{#lex=auction] NP ) ) : % { $0:$0, $1 :$1, $2:$2, $6:$4); # default %(#lex : #lex)" | |
| }, | |
| "TABREF2": { | |
| "type_str": "table", | |
| "html": null, | |
| "content": "<table/>", | |
| "text": "Example of some discrepancies", | |
| "num": null | |
| }, | |
| "TABREF3": { | |
| "type_str": "table", | |
| "html": null, | |
| "content": "<table/>", | |
| "text": "Selected Structural Dictionary Entries", | |
| "num": null | |
| }, | |
| "TABREF5": { | |
| "type_str": "table", | |
| "html": null, | |
| "content": "<table/>", | |
| "text": "Selected Bilingual Dictionary Entries", | |
| "num": null | |
| } | |
| } | |
| } | |
| } |