{ "paper_id": "W98-0106", "header": { "generated_with": "S2ORC 1.0.0", "date_generated": "2023-01-19T06:04:48.853756Z" }, "title": "Can the TAG derivation tree represent a semantic graph? An ans wer in the light of Meaning\u2022 Text Theory", "authors": [ { "first": "Marie-Helene", "middle": [], "last": "Candito", "suffix": "", "affiliation": { "laboratory": "", "institution": "Universile Paris", "location": { "addrLine": "7 2, place Jussieu", "postCode": "7003, 75251", "settlement": "Paris cedex 05 marie" } }, "email": "" }, { "first": "Sylvain", "middle": [], "last": "Kahane", "suffix": "", "affiliation": { "laboratory": "", "institution": "Universile Paris", "location": { "addrLine": "7 2, place Jussieu", "postCode": "7003, 75251", "settlement": "Paris cedex 05 marie" } }, "email": "" } ], "year": "", "venue": null, "identifiers": {}, "abstract": "", "pdf_parse": { "paper_id": "W98-0106", "_pdf_hash": "", "abstract": [], "body_text": [ { "text": "From the parsing point of view, the derivation tree in TAG [hereafter DT] is seen as the \"history\" of the derivation but also as a linguistic representation, closer to semantics, that can be the basis of a further analysis. Because in TAG the elementary trees are lexicalized and localize \u2022 the predicate-arguments relations, several works have compared the DT to a structure involving dependencies between lexical items (RJ92; RVW95). 1 We agree with these authors that there are divergences between the DT and syntactic dependencies, but we show here that the DT -in the sense of (SS94) -can be viewed as a semantic dependency graph, namely a Sems for Meaning-Text Theory (MTI] (ZM67; M88) . This requires the predicate-argument cooccurrence principle and also constraints on the adjunction of prcdicative auxiliary Lrees. We briefly introduce the representation levels in MTI before studying the dependencies shown by the DT. 2", "cite_spans": [ { "start": 436, "end": 437, "text": "1", "ref_id": null }, { "start": 674, "end": 679, "text": "(MTI]", "ref_id": null }, { "start": 680, "end": 686, "text": "(ZM67;", "ref_id": "BIBREF14" }, { "start": 687, "end": 691, "text": "M88)", "ref_id": "BIBREF8" } ], "ref_spans": [], "eq_spans": [], "section": "Introduction", "sec_num": null }, { "text": "MTI distinguishes between linguistic representations and correspondance rules to go from a representation to another, at an adjacent level. For a wr\u00fcten sentence, there are 5 representations, each with a central structure : semantic [SemS] , deep and surface syntactic [DSyntS and SSyntS] , deep and surface morphological [DMorphS and SMorphS] . At each Ievel, additional structures may supplement the central structure. A key feature of MTI is that it distinguishes between semantic and syntactic dependencies. The Sems is a graph showing semantic dependencies between semantemes (= semantic units). The dependencies are numbered to distinguish between the different 1 (RJ92) relate the DT to the deep syntactic structure (DSyntS) of MTI, namely a syntactic dependency tree, but they note thut this correspondence DT ! DSyntS is not direct, bccause the interpretation of adjunction arcs in terms of dependencies is not constant. (RVW95) take this divergence between DT and dependency tree as one of the motivations for defining D-Tree Grammars. 2 We are thankful to Anne Abei\\le, Laurence Danlos and Owen Rambow for valuable comments on earlier versions of this work.", "cite_spans": [ { "start": 233, "end": 239, "text": "[SemS]", "ref_id": null }, { "start": 269, "end": 288, "text": "[DSyntS and SSyntS]", "ref_id": null }, { "start": 322, "end": 343, "text": "[DMorphS and SMorphS]", "ref_id": null }, { "start": 1046, "end": 1047, "text": "2", "ref_id": null } ], "ref_spans": [], "eq_spans": [], "section": "Representation levels in MTT", "sec_num": "1." }, { "text": "arguments of a predicative Semanteme. An atlthe DT and SemS f\u00fcrsentence ( l ). 9subslitution site treesemamic-govemoradjunclion site cn:esemanlic de~ndent11 1 1 substituted=J semanlicadjoined= 1 semantictreedeoendenttn:.:gov.:mor), wheremultiple modifier adjunctions are allowcd at thesame address. 101 The fact that the DT should represent semantics isnot new. See for example (A93) who distinguishesbetween g/ass in a wine glass anu in a glass of wine onpurely semantic grounds; (K89) who mentions <hatTAG should \"preserve a straightforward compositionalsemantics\"; (098) who descirbesG-TAG. ageneration system based on TAG where a derivationtree is built by lexicalizing a conceptual structurc.9 The TAG analysis is from (X95), except thatdetenniners arenot consideredasnominalcomplements and are thus adjoined." } } } }