| { | |
| "paper_id": "1991", | |
| "header": { | |
| "generated_with": "S2ORC 1.0.0", | |
| "date_generated": "2023-01-19T07:35:35.492316Z" | |
| }, | |
| "title": "Using inheritance in Object-Oriented Programming to combine syntactic rules and lexical idiosyncrasies", | |
| "authors": [ | |
| { | |
| "first": "Benoit", | |
| "middle": [], | |
| "last": "Habert", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "Ecole Nortnale Superieure de Fontehay Saint Cloud", | |
| "location": { | |
| "addrLine": "31 avenue Lomba.rt", | |
| "postCode": "F-92260", | |
| "settlement": "FONTENAY-AUX-ROSES", | |
| "country": "FRANCE" | |
| } | |
| }, | |
| "email": "" | |
| } | |
| ], | |
| "year": "", | |
| "venue": null, | |
| "identifiers": {}, | |
| "abstract": "In parsing idioms and frozen expressions in French, ohe needs to \u2022 combine general syntactic rules and idiosyncratic-constraints. The inheritance structure \u2022 provided by Object Oriented Programming languages, and more spedf.ically the combination of methods present in ct: bs, Common Lisp Gbjed System, \u2022 app ears-as an elegant ; and-efficient , approach to ' deal with such a complex interaction-. in parsiIYg hHom' s-antl frozen-expre' sst' on: s 1n Fren\ufffdi\\\" one n: e' ei:l ' s to \u2022 comb1ne g' enel a l : syntacH:c rules a\u2022 na idlosyncra.Ht ' constrai1nls. As : a mat' fet \u2022 of 'i a t:l; representing such an interacHon vla an inheritance laHke \u2022 appears , a:s \u2022 an elegah t and : \ufffdflkientapproach. For t!1e s ake of explanation; EngHsh-Mtoms wi-11 he used as exampM,-. However this t\ufffdmbh1ift-ion \u2022 of synlacHc rules and idi-osyn:cr-atk ' behaviour via manipulations bf the tnherH:ance struftu:re and lhe methods artad1ed to u> has beendesigned for French cbmpound adverbials. More lhan 6,boo compou.ntl \u2022 adverbial's have be' eii Hsred and ' s\u00b0f udie t l il' t t!i\\fit 1 f (Gross, 19' 9' 0). A lexicon gtammat \u2022 \u2022 ot \ufffd,525 \u00b0compoun<1 advetb1als coming from the tAbt files has been used 1n ; parsing a test c cotpus of 72;000 words. toiOMS: A-PECULIAR COMBINATION OF REctutARitiES AND lDIOSYNCRASIES The semantics of idioms will hot he-accounted fbt here, since it is a cohftovetsial", | |
| "pdf_parse": { | |
| "paper_id": "1991", | |
| "_pdf_hash": "", | |
| "abstract": [ | |
| { | |
| "text": "In parsing idioms and frozen expressions in French, ohe needs to \u2022 combine general syntactic rules and idiosyncratic-constraints. The inheritance structure \u2022 provided by Object Oriented Programming languages, and more spedf.ically the combination of methods present in ct: bs, Common Lisp Gbjed System, \u2022 app ears-as an elegant ; and-efficient , approach to ' deal with such a complex interaction-. in parsiIYg hHom' s-antl frozen-expre' sst' on: s 1n Fren\ufffdi\\\" one n: e' ei:l ' s to \u2022 comb1ne g' enel a l : syntacH:c rules a\u2022 na idlosyncra.Ht ' constrai1nls. As : a mat' fet \u2022 of 'i a t:l; representing such an interacHon vla an inheritance laHke \u2022 appears , a:s \u2022 an elegah t and : \ufffdflkientapproach. For t!1e s ake of explanation; EngHsh-Mtoms wi-11 he used as exampM,-. However this t\ufffdmbh1ift-ion \u2022 of synlacHc rules and idi-osyn:cr-atk ' behaviour via manipulations bf the tnherH:ance struftu:re and lhe methods artad1ed to u> has beendesigned for French cbmpound adverbials. More lhan 6,boo compou.ntl \u2022 adverbial's have be' eii Hsred and ' s\u00b0f udie t l il' t t!i\\fit 1 f (Gross, 19' 9' 0). A lexicon gtammat \u2022 \u2022 ot \ufffd,525 \u00b0compoun<1 advetb1als coming from the tAbt files has been used 1n ; parsing a test c cotpus of 72;000 words. toiOMS: A-PECULIAR COMBINATION OF REctutARitiES AND lDIOSYNCRASIES The semantics of idioms will hot he-accounted fbt here, since it is a cohftovetsial", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Abstract", | |
| "sec_num": null | |
| } | |
| ], | |
| "body_text": [ | |
| { | |
| "text": "problem. LFG, GPSG and TAGs made quite different claims on this topic 2 . Within a syntactic category, it has been shown for French, at LADL, that frozen expressions are generaily mote numerous than 'free' ones: 20, 000 frozen verbs (12,000 free), 6,000 adverbials (1,500 free). More than \u2022 25, UOO compound nouns have been studied so far, but their number is far greater, as they -consHtute the fna, jor part of new termsin subia\ufffdgttages (Grishman ' & l<ittred.ge, 1986) While foflow. ttr g ro a tatge \ufffdxtent the : gen\ufffdtal \u2022 syntatHc \"f ul\ufffd:s, . ftijzen ' expre'ssi' ons present : itiiosyrforas1es. At a : sy n' tat :tlt \"l' evel, -an id.ibm can acce . pt \u2022 a moaHie-r _ (\"in (-loving) memory of'), or not (#'by the new way\") 3 . It can require certain syntactic. features for some of its constituents. For instance, it may need a certain type of determiner: \"for the sake of\" versus \"#for a sake of\". Lastly, an idiom is associated with \u2022 fixed lexical items. Usually it is not possible to replace them by synonyms: #\"by the road\" versus \"by the way\". Since most of frozen expressions follow general syntactic rules, and since 'transformations' apply to them, it is not reasonable to try and process them in a first lexical step. Recognizing idioms belongs therefore to the whole syntactic analysis. Nevertheless their idiosyncratic features must be taken into account in rules.", | |
| "cite_spans": [ | |
| { | |
| "start": 439, | |
| "end": 471, | |
| "text": "(Grishman ' & l<ittred.ge, 1986)", | |
| "ref_id": null | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "OLMES 4 is a general parser written in CLOS 5 (Keene, 1989; Steele, 1990, p 770-864) , and \u2022tested with the Victoria Day implementation of .PCL 6 (provided by Xerox Labor_ atories), using Lucid Common Lisp 3.0.1, oil. a Sun 3 workstation, at LITP 7. OLMES belongs to the active chart parser family. The inpunext can be parsed from left to right, or the other way round, or even both ways at the same time (around pivots). Top-down, bottom up or bottom-up then top-down strategies are available. The rules used by OLMES: follow the formalism created for P ATR-II (Shieber, 1986), because it is a kind of \"lingua franca \" fo r unification-based grammars. Additional constraints can be associated with ordinary context-free rules so as to analyse mildly context-sensitive languages (Gazdar, 1988) . Each symbol in the rule is the root of a Directed Acyclic Graph (DAG). \u2022 In such category structures, each edge is labelled, and leads either to an atom or to another.complex category structure- (Gazdar et al., 1988 The sequence of a first right-hand side symbol dominating a DAG with an edge \"cat(egory)\" having \"prep(osition)\" as its value, a second symbol with \"cat\" \"det(erminer)\", and a third symbol with \"noun\" as \"cat\" makes an \"adv(erbial)\". Additionally the second and the third symbol must share the same value for the feature \"agreement\". The rule above would recognize as idioms \"at the moment\", \"in a way\", \"in the end\", using this toy lexicon. Note that the completed rule is more restrictive than the context-free part of it. The latter would accept \"*by those way\", the former would not, because \"those\" and \"way\" do not agree.", | |
| "cite_spans": [ | |
| { | |
| "start": 46, | |
| "end": 59, | |
| "text": "(Keene, 1989;", | |
| "ref_id": "BIBREF12" | |
| }, | |
| { | |
| "start": 60, | |
| "end": 84, | |
| "text": "Steele, 1990, p 770-864)", | |
| "ref_id": null | |
| }, | |
| { | |
| "start": 779, | |
| "end": 793, | |
| "text": "(Gazdar, 1988)", | |
| "ref_id": "BIBREF4" | |
| }, | |
| { | |
| "start": 991, | |
| "end": 1011, | |
| "text": "(Gazdar et al., 1988", | |
| "ref_id": "BIBREF7" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "STATING THE GENERAL BEHAVIOUR OF A FAMILY OF IDIOMS", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "prep det noun", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "There are two main classes: active and inactive. An inactive agent corresponds to a (possibly partial) constituent which has been found. For instance, for each left-hand side symbol in the grammar, a class is created inheriting from the inactive agent class. The active agents correspond to the right-hand side symbols of the grammar . . Each of them is searching for a constituent meeting certain constraints, as defined in the corresponding DAG in the rule. If it finds such a constituent, it then creates an instance of the class corresponding to the following symbol in the right-hand side part of the rule. When the last active agent of the rule \"succeeds\", it creates an instance of the class corresponding to the left-hand side of the rule. The pivot of the rule is the symbol starting the whole analysis. It need not be the left-most one. For the -rule above, in bottom-up parsing, four classes are defined: LH S-1, RHSl-2, RHS2-3, RHS3-4, respectively (figure 2). RHS l-2, RHS2-3 and RHS 3-4 are subclasses of LHS-1, their instances will be active agents examining the text from right to left. The pivot of the rule is the class RH s 3 -4 (in bold font), corresponding to a noun.", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "OLMES creates an instance of RH S3-4 . This instance searches for a noun, and finds it: \"moment\". It creates an instance of RH s 2 -3 which examines the word on the left of \"moment\", and which stores a partial parse tree. If this word is a determiner, and has a feature \"agreement\" matching with the corresponding feature of \"moment\", the new partial parse tree is transmitted to the instance of \ufffdHSl-2 which is then created and whose constraints are matched against the word on the left of the determiner found by the instance of RH s 2 -3. In the case that \u2022 the instance of RHSl-2 finds a preposition, it then creates an instance of LH s -1 storing the complete parse tree and the additional information gathered from the unification on the rest of the DAGs.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "To indicate that a word can belong to the type of idiom described in the rule, the lexicon associates the class-name RHS 3-4 with this word. It could be the case for the word \"moment\". In a bottom-up analysis, for each occurrence of \"moment\" in the input text,", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "Changing the grammar rules from sequences of 'passive' labels to a network of active classes makes it possible to increase as necessary the knowledge the instances of these classes can utilise, and to use inheritance not only in the lexicon (Shieber, 1986), but in the grammar rules as well.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "To indicate that a word can belong to the type of idiom described in the rule, the lexicon associates the class-name RHS 3-4 with this word. It could be the case for the word \"moment\". In a bottom-up analysis, for each occurrence of \"moment\" in the input text,", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": ".", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "USING THE .\u2022 iNHERITANCE STRUCTURE", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": ".", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "USING THE .\u2022 iNHERITANCE STRUCTURE", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "The rule stated above is not restrictive enough. For instance, it would parse as an idiom \"by a way\" in the sentence: \"he arrived by a way new to me\". It would be rather an unsatisfactory approach to\u2022 create as many rules as combinations found between the . ", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "TO TAKE IDIOSYNCRASIE S INTO ACCOUNT", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "adv prep det One of the salient characteristics of CLOS, inherited from its ancestors, COMMON LOOPS (Bobrow et al., 1986) and NEW FLAVORS, is the control given over the combination of methods having the same name and present in the super-classes of a given class (Keene, 1989) When coming across \"way\" in the input text, OLMES would therefore create one instance of each class. For example, in the case of the instance of adv = prep_de f ini te-a rt i c 1 e_no un, because this class is a subclass of <let-definite-article, and because the method combination for cons t rain t s is redefined, the constraints inherited via det-defini te-article and <let-article are added to the general constraints defined in the rule and inherited through RHS3-4. The arguments following the name of the class are used as well. In the end, the parser will actually try the following rule (figure 8):", | |
| "cite_spans": [ | |
| { | |
| "start": 87, | |
| "end": 121, | |
| "text": "COMMON LOOPS (Bobrow et al., 1986)", | |
| "ref_id": null | |
| }, | |
| { | |
| "start": 263, | |
| "end": 276, | |
| "text": "(Keene, 1989)", | |
| "ref_id": "BIBREF12" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "lhs rhs 1 rhs2 rhs3", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "LHS -> RHSl RHS2 RHS3 <LHS cat> = adv\u2022 <RHSl cat> = prep <RHSl form> = <RHS3 prepl form> <RHS2 cat> = det <RHS3 . cat> = noun This rule will accept \"by the way\", but will reject ''by a way\", \"in the way\" ... The rule and the parameters for \"moment\" would allow the parsing of \"at the moment\", and those f<;>r \"opinion\" the acceptance of \"in my opinion\" ... ", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "lhs rhs 1 rhs2 rhs3", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "Recent work, within the TAG formalism (Abeille, 1990; Abeille & Schabes, 1989) , claimed that idioms should be parsed during the whole syntactic analysis, using the same formal devices as for parsing non idiomatic expressions. This approach uses a slightly modified version of T AGs, namely lexicalized T AGs, in which each 'rule', i.e. each tree, is anchored with a lexical As a matter of fact, giving the rules a pivot and associating the words in the lexicon with these pivots\ufffd as shown above, is a first step in lexicalizing a grammar. On the other hand, Lexicalized T AGs do not use phrase structure rules any more, but trees directly stating to any depth the constituents needed, their structure, and possibly their lexical heads. For this very reason, the TAG formalism deals with idioms in a more natural and powerful way. For the sake of explanation, the rules given in this paper are flattening the structure of the phrases. In order to give to the relevant idioms the same structure as the corresponding free phrases, one would need some complex transmission of features among related rules (Habert, 1991) .", | |
| "cite_spans": [ | |
| { | |
| "start": 38, | |
| "end": 53, | |
| "text": "(Abeille, 1990;", | |
| "ref_id": "BIBREF0" | |
| }, | |
| { | |
| "start": 54, | |
| "end": 78, | |
| "text": "Abeille & Schabes, 1989)", | |
| "ref_id": "BIBREF1" | |
| }, | |
| { | |
| "start": 1103, | |
| "end": 1117, | |
| "text": "(Habert, 1991)", | |
| "ref_id": null | |
| } | |
| ], | |
| "ref_spans": [ | |
| { | |
| "start": 305, | |
| "end": 375, | |
| "text": "AGs, in which each 'rule', i.e. each tree, is anchored with a lexical", | |
| "ref_id": null | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "RELATED WORK: PARSING IDIOMS IN TREE ADJOINING GRAMMARS (TAGS)", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "Relying to such an extent on the inheritance structure partly breaks the decentralization rule which is central to object oriented programming 11. When slightly modifying a class, here is a risk of triggering a 11 (Meyer, 1988, p 251) In most cases, clients of a class should not need to know the inheritance structure that led to its implementation.", | |
| "cite_spans": [ | |
| { | |
| "start": 214, | |
| "end": 234, | |
| "text": "(Meyer, 1988, p 251)", | |
| "ref_id": null | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "STRUCTURING GRAMMARS VIA INHERITANCE", | |
| "sec_num": null | |
| } | |
| ], | |
| "back_matter": [ | |
| { | |
| "text": "chain reaction of changes. As Sakkinen, 1989, states:Features aiming at \"exploratory programming\" need not necessarily make the programmer into a Vasco de Gama or an Amundsen; (s)he may well become Alice in Wonderland, never knowing what metamorphoses some seemingly innocent act may cause. The danger is a real one. Nevertheless, so far, it has been most beneficial to take advantage of the inheritance structure to portray the linguistic knowledge we are dealing with. In doing so, we stress the classification tools present in Objet-Oriented Programming Languages: the inheritance lattice is used to progressively constrain the class of the solution (Wegner, 1987) . This approach uses a unification-based formalism with a clear-cut distinction between phrase structure rules and subcategorization frames. In spite of this, it combines properly the generalizations stated by the syntactic rules and additionnal constraints necessary to account for the idiosyncrasies that the idioms show. This solution is by no means limited to frozen expressions. It contributes to a clear expression of the complex interactions found in the grammar between syntactic and lexical rules (Abeille 90). It is thus worth investigating the ways in which inheritance can help in structuring not only the lexicon but also the grammar. ", | |
| "cite_spans": [ | |
| { | |
| "start": 653, | |
| "end": 667, | |
| "text": "(Wegner, 1987)", | |
| "ref_id": null | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "annex", | |
| "sec_num": null | |
| } | |
| ], | |
| "bib_entries": { | |
| "BIBREF0": { | |
| "ref_id": "b0", | |
| "title": "Lexical and syntactic rules in a tree adjoining grammar", | |
| "authors": [ | |
| { | |
| "first": "Abeille", | |
| "middle": [], | |
| "last": "Anne", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1990, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Abeille Anne 1990 \"Lexical and syntactic rules in a tree adjoining grammar\", ACL'90", | |
| "links": null | |
| }, | |
| "BIBREF1": { | |
| "ref_id": "b1", | |
| "title": "Parsing Idioms in Lexicalized TAGs", | |
| "authors": [ | |
| { | |
| "first": "Abeille", | |
| "middle": [], | |
| "last": "Anne", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Yves", | |
| "middle": [], | |
| "last": "Schabes", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1989, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Abeille Anne, Yves Schabes 1989 \"Parsing Idioms in Lexicalized TAGs\", EACL'89.", | |
| "links": null | |
| }, | |
| "BIBREF2": { | |
| "ref_id": "b2", | |
| "title": "The mental representation of grammatical representation", | |
| "authors": [], | |
| "year": 1982, | |
| "venue": "The passive in Lexical Theory", | |
| "volume": "87", | |
| "issue": "", | |
| "pages": "2--86", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Bresnan Joan 1982a (editor) The mental representation of grammatical representation, The MIT Press. 87 1982b \"The passive in Lexical Theory\", (Bresnan, 1982a, p 2-86).", | |
| "links": null | |
| }, | |
| "BIBREF4": { | |
| "ref_id": "b4", | |
| "title": "Applicability of Indexed Grammars to natural languages", | |
| "authors": [ | |
| { | |
| "first": "Gazdar", | |
| "middle": [], | |
| "last": "Gerald", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1988, | |
| "venue": "Natural language parsing and linguistic theories. Reyle and Rohrer", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Gazdar Gerald 1988 \"Applicability of Indexed Grammars to natural languages\", in Natural language parsing and linguistic theories. Reyle and Rohrer editors, D. Reidel Publishing Company.", | |
| "links": null | |
| }, | |
| "BIBREF5": { | |
| "ref_id": "b5", | |
| "title": "Ivan Sag 1985 Generalized Phrase Structure Grammar", | |
| "authors": [ | |
| { | |
| "first": "Gazdar", | |
| "middle": [], | |
| "last": "Gerald", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Ewan", | |
| "middle": [], | |
| "last": "Klein", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Geoffrey", | |
| "middle": [], | |
| "last": "Pullum", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": null, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Gazdar Gerald, Ewan Klein, Geoffrey Pullum, Ivan Sag 1985 Generalized Phrase Structure Grammar. Harvard University Press.", | |
| "links": null | |
| }, | |
| "BIBREF6": { | |
| "ref_id": "b6", | |
| "title": "Mellish Chris 1989 Natural Language Processing in Lisp", | |
| "authors": [ | |
| { | |
| "first": "Gazdar", | |
| "middle": [], | |
| "last": "Gerald", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": null, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Gazdar Gerald, Mellish Chris 1989 Natural Language Processing in Lisp. Addison-Wesley.", | |
| "links": null | |
| }, | |
| "BIBREF7": { | |
| "ref_id": "b7", | |
| "title": "Category structures", | |
| "authors": [ | |
| { | |
| "first": "Gazdar", | |
| "middle": [], | |
| "last": "Gerald", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Geoffrey", | |
| "middle": [ | |
| "K" | |
| ], | |
| "last": "Pullum", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Robert", | |
| "middle": [], | |
| "last": "Carpenter", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Ewan", | |
| "middle": [], | |
| "last": "Klein", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Thomas", | |
| "middle": [ | |
| "E" | |
| ], | |
| "last": "Hukari", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Robert", | |
| "middle": [ | |
| "D" | |
| ], | |
| "last": "Levine", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1988, | |
| "venue": "Computational Linguistics", | |
| "volume": "14", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Gazdar Gerald, Geoffrey K. Pullum, Robert Carpenter, Ewan Klein, Thomas E. Hukari, Robert D. Levine 1988 \"Category structures\", Computational Linguistics. Vol. 14, #1.", | |
| "links": null | |
| }, | |
| "BIBREF8": { | |
| "ref_id": "b8", | |
| "title": "Analyzing language in restricted domains: sublanguage description and processing", | |
| "authors": [ | |
| { | |
| "first": "Grishman", | |
| "middle": [], | |
| "last": "Ralph", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Richard", | |
| "middle": [], | |
| "last": "Kittredge", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1986, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Grishman Ralph, Richard Kittredge (editors) 1986 Analyzing language in restricted domains: sublanguage description and processing. Lawrence Erlbaum Associates.", | |
| "links": null | |
| }, | |
| "BIBREF9": { | |
| "ref_id": "b9", | |
| "title": "Sur les phrases figees complexes du fran\ufffdais", | |
| "authors": [ | |
| { | |
| "first": "Maurice", | |
| "middle": [], | |
| "last": "Gross", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1988, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Gross Maurice 1988 \"Sur les phrases figees complexes du fran\ufffdais\", Langue Francaise 77.", | |
| "links": null | |
| }, | |
| "BIBREF10": { | |
| "ref_id": "b10", | |
| "title": "Controlling the generic dispatch to represent domain knowledge", | |
| "authors": [ | |
| { | |
| "first": "Habert", | |
| "middle": [], | |
| "last": "Benoit", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1990, | |
| "venue": "Proceedings of the third CLOS users and implementors workshop. OOPSLA'90", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Habert Benoit 1990 \"Controlling the generic dispatch to represent domain knowledge\", Proceedings of the third CLOS users and implementors workshop. OOPSLA'90.", | |
| "links": null | |
| }, | |
| "BIBREF11": { | |
| "ref_id": "b11", | |
| "title": "Lan gage s a objets et an a 1 y s e linguistigue", | |
| "authors": [], | |
| "year": null, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Lan gage s a objets et an a 1 y s e linguistigue. Doctoral thesis.", | |
| "links": null | |
| }, | |
| "BIBREF12": { | |
| "ref_id": "b12", | |
| "title": "Object-Oriented Programming in Common Lisp", | |
| "authors": [ | |
| { | |
| "first": "Keene", | |
| "middle": [], | |
| "last": "Sonya", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "E", | |
| "middle": [], | |
| "last": "", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1989, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Keene Sonya E. 1989 Object-Oriented Programming in Common Lisp. Addison Wesley.", | |
| "links": null | |
| }, | |
| "BIBREF13": { | |
| "ref_id": "b13", | |
| "title": "Parsing in Functional Unification Grammar", | |
| "authors": [ | |
| { | |
| "first": "Kay", | |
| "middle": [], | |
| "last": "Martin", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1985, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Kay Martin 1985 \"Parsing in Functional Unification Grammar\", in Natural language parsing, D. Dowty, L. Karttunen and A. Zwicky editors, Cambridge University Press.", | |
| "links": null | |
| } | |
| }, | |
| "ref_entries": { | |
| "FIGREF0": { | |
| "uris": null, | |
| "num": null, | |
| "text": ". A small propotHon ot frozen expressions have constitnen'f s , exi-snng only in such \u2022 conlex'ts (snch as \"umbra ge'\u2022 ,n \u2022 \u2022\ufffdte tak\ufffd um'o rage at NP -. \ufffd' i -or : \ufffdrte .' ta\ufffdn , from re>reign : ianguag\ufffds (''a pri0rf i \u2022t \u2022or t-oUbw letS-citl:en . rules\u2022 . Apa.ft from these matgina: t ' case$; l\ufffdloms consis't ' of the same words as fwe i t.re\ufffd phrases, and tlrey Ifiriow tl\\'e same syntactit : rni@s\ufffd \u2022 \u2022\ufffdm \u2022 contrast\\ ...-,,Y the way' \\ for ln.-slance, are Jtist ordinarf PP; FutHl' ermbte ) a\u2022 s shoWh _ tot Engitsh by Wa\u2022 sc>' w et ; at 1982-and -fb'r -Frenih by \u2022 \u2022 Gto\u2022 ss-. , 1\"88, Woo. 't he \u00a7vnt:atlic behaviour , o \u2022 -f Idioms 1s . fh: uch :f nore s_ y-s t emaHt than \u2022 is us11aRy thought: 'lra-nsfotmaHo't ,' s\u2022\u2022 \u2022 apply fo them. & >me tdn :tl ' of \u2022 \u2022m\ufffdtaihll-\ufffds' mu. \u2022 st he th'en te seci :to atcoun:t for the s e -t\ufffdiated \u2022 struchi-t es.", | |
| "type_str": "figure" | |
| }, | |
| "FIGREF1": { | |
| "uris": null, | |
| "num": null, | |
| "text": "preposition and the type of determiner used \u2022 in such idioms. What we need .instead is a means to adjoin new constraints to the set of conditions \u2022 defined . in the rule, in a modular way, that is, using inheritance. In the CLOS philosophy, if means that some 'mixin' classes are created. Such classes are not intended to have instances on their own. On the contrary, they are only used as constituents (super classes) in defining more specialized classes. For instance, one can define the following'mixin' classes (see figure 3). Each 'mixin' class used to specialize the rule has a method constraints which states particular constraints on the determiner. The content of this method (in PA TR form) follows the class nameSome classes for the constraifits on determiners The rule given above (figure 1) is slightly redefined : from now on, the pivot transmits to the RHSl the form of preposition, and to the RHS2 precisions on the type of determiner which is needed (the dark nodes indicate this sharing of values in figure 4).", | |
| "type_str": "figure" | |
| }, | |
| "FIGREF2": { | |
| "uris": null, | |
| "num": null, | |
| "text": "p re p _ definite-a rti c I e _noun", | |
| "type_str": "figure" | |
| }, | |
| "FIGREF3": { | |
| "uris": null, | |
| "num": null, | |
| "text": "noted that the constraints on the preposition and the conditions on the determiner are not on the same level. The latter are in a way more syntactic: some syntactic properties of the idiom as a whole depend on the nature of the determiner. The form of the initial preposition is purely idiosyncratic. It does not even always contribute to the meaning of the expression. For that reason, the way to specify the initial preposition does not use inheritance. A list of associations { <parameter> <value>} is being used at the initialization of the instance of a given pivot class to deal with such litteral constraints. For instance, the list \"RHSl by\" will trigger the adjunction of: same reason). The last one (in bold font) shadows the others. 9 We do not use the actual Lisp syntax for the result, as it is not relevant. Standard method combination", | |
| "type_str": "figure" | |
| }, | |
| "FIGREF4": { | |
| "uris": null, | |
| "num": null, | |
| "text": "1 O . In this case, it is . possible to specify that all the methods constraints accessible from a given class should be called in turn, and the final result should be the addition of all the returned values. With this combination of methods, the result of the function call ( constraints det-definite-article-1 ) would be (figure 7): <RHS3 detl cat-precisions determiner-type> = article 10 When the most specific method represents nothing but an \u2022 \u2022addition to the action of one super method, it is generally possible in an Object Oriented Programming Language to combine it with this super-method, so as to share common behaviours. \".'\ufffdtra:i\ufffdt\ufffd . \u2022 \u2022 <instance\u2022 of \u2022 : Oet: de:t {njt:e \ufffd article>\") ...... \u2022 -\u2022 .. Special method combinationThis method combination provides a means to add constraints present in the inheritance lattice, and only the relevant ones. In the lexicon, the entries for pivots of ad verbials could mention (among other information):", | |
| "type_str": "figure" | |
| }, | |
| "FIGREF6": { | |
| "uris": null, | |
| "num": null, | |
| "text": "Sequence of DAGs for the-final ruleThis very simple example does not do justice to the complexity of syntactic and lexical properties of adverbial idioms. However it stresses the hierarchy of these features, and the way in which the inheritance graph can, at the same time, _ mirror this structure and take advantage of it. Note that the 'mixin' classes defined above are useful as such. They constitute the primitives to complete the basic syntac;tic rules. They correspond to organized constraints which are interesting on their own, . as they can be reused in different contexts.\u2022 For instance, another rule for ad verbials is:LHS -> RH Sl RHS2 .RHS3 RHS4. <LHS cat> be created for each symbol of the rule. For example, the class corresponding to the pivot (the noun) is RHS4-9. New specialized classes are then defined: \"with an eye to\" and \"in the form of'', respectively. It is possible in OLMES to express that a certain word can enter different linked syntactic structures at the same time, thus providing 'meta-rules'.One of theses families of rules is the class:[adv = prep definitethe correct values for the parameters RH Sl and RHS4, possibly leading to the parsing of for the sake of or for its sake.", | |
| "type_str": "figure" | |
| }, | |
| "FIGREF7": { | |
| "uris": null, | |
| "num": null, | |
| "text": "item. In fact, there are no separate phrase structure rules any more: they are collapsed into the lexicon. Only the relevant rules are used while parsing, as they are triggered by lexical items.Meta-rules are provided by means of families of trees. The trees corresponding to idioms include several lexical items: take and bucket in the case of to take the bucket . As an additional filtering, the search of an idiom is triggered only if all the lexical heads are actually present in the sentence, in the right order\ufffd", | |
| "type_str": "figure" | |
| }, | |
| "TABREF0": { | |
| "type_str": "table", | |
| "num": null, | |
| "content": "<table/>", | |
| "text": "", | |
| "html": null | |
| }, | |
| "TABREF1": { | |
| "type_str": "table", | |
| "num": null, | |
| "content": "<table><tr><td>A graphical equivalent could be: lhs rhs 1 rhs2 caI c:t I adv . . prep det Figure 1: momen t cat noun agreement number singular my cat det cat-precisions determiner-type possessive the cat det cat-precisions determiner-type article article-type de finite this cat det cat precisions determiner-type demonstrative those cat det cat-precisions determiner-type demonstrative agreement number plural way cat noun agreement number singular</td><td>rhs3 .noun cat</td></tr></table>", | |
| "text": "", | |
| "html": null | |
| } | |
| } | |
| } | |
| } |