REALPRO General English Grammar User Manual
Contents
1. fem E mut pam O dual teacher Specification gender dual should yield pronominal choices such as his or her but is not currently implemented 9 NOUNS 20 e Feature case can have the following values gen beans her Fob bem him This feature is not usually used in an input DSyntS for REALPRO e Definite indefinite and demonstrative determiners can be introduced through features Feature article can have the following values dd the tiara eta y dewprox this ars these Haras BEEN that tiara 9 2 Example First example Noun the yemen dss OUTPUT The Yemen END DSYNTS Yemen class proper_noun article def END Second example 9 NOUNS 21 fA ii cesses Some tiaras p Sane as eae c c C DSYNTS tiara class common noun number pl END Third example OUTPUT These cars END DSYNTS car class common noun article dem prox number pl END 9 3 Notes e The feature combination article indef number pl yields the bare plural tiaras Note that since article indef is the default sim ply number pl also yields the bare plural To obtain some as a determiner use lexeme SOME as an ATTR to the noun Note that SOME does not have a number so to obtaon some tiaras you need to indicate number pl on the noun See Noun some duck dss and Noun some ducks dss e In English the distinction between dative and2. DSYNTS eat class verb tense pres I John class proper noun II bean class common noun number pl article no art ATTR often class adverb END 18 ADJUNCTS TO A CLAUSE 44 Second example Lp SERES a SS pe ea an Often John eats beans DSYNTS eat class verb tense pres I John class proper noun II bean class common noun number pl article no art ATTR often class adverb starting point END Third example A SS E e John eats beans often DSYNTS eat class verb tense pres I John class proper noun II bean class common noun number pl article no art ATTR often class adverb rheme END Fourth example OUTPUT If you had money I would do anything for you 18 ADJUNCTS TO A CLAUSE 45 END DSYNTS DO mood cond I 1 class personal_pronoun number sg person 1st II anything class partitive_pronoun person 3rd number sg ATTR FOR1 II you class personal pronoun number sg person 2nd ATTR HAVE1 tense past starting point I you class personal pronoun number sg person 2nd II money class common noun article no art ATTR IF END 18 3 Notes e The positioning of the adjunct to a clause is independent of the type of adjunct adverbial prepositional clausal The default position im mediately pre verbal is not alwa
3. class verb II give class verb I Mary class proper noun gender fem pro wh III John class proper noun pro wh gender masc rheme II book class common noun number pl article no art I authority class common noun number pl article def END 17 3 Notes e The grammar automatically determines the need for an auxiliary based on the grammatical function of the fronted wh word and the embed ded matrix status of the verb It also fronts at most one wh word 18 ADJUNCTS TO A CLAUSE 42 17 4 Shortcomings e Adjuncts are currently not handled only arguments labeled I II or HI e Also prepositions present at DSyntS are not handled even if they are marked as an argument and their argument is marked pro wh e Genitive wh words whose are not handled e For echo questions with wh words in situ such as You gave books to whom do not use the pro wh feature but instead specify the word as a noun and add a question mark see Section 22 page 52 18 Adjuncts to a Clause 18 1 Description Adjuncts to a clause such as adverbs adverbial phrases prepositional phrases and adjunct clauses are related to the verb they modify by the ATTR rela tionship There are three positions for adjuncts sentence initial immediately pre verbal and sentence final As a default prepositional and clausal adjuncts appear in sentence final position John ate beans while waiting for Nancy while adverbial adjuncts appe
4. colon precedes dash precedes comma e Adding feature rightmost punct colon to the root node will not have any effect since the point absorption mechanism will favor the period over the colon Instead use feature end punct colon e The system automatically transposes quotes and periods in sentence final position following standard convention 22 4 Shortcomings e Point absorption also happens with brackets and parentheses He was nice but slow and I thanked him 23 HTML ANNOTATIONS 95 23 HTML Annotations 23 1 Description To add an HTML annotation tag with or without attributes proceed as follows e To add an HTML tag o to the string corresponding to just the node in question add sgml o where o can be any HTML tag eg A for anchor B for bold etc e To add an HTML tag o with attributes 5 to the string corresponding to just the node in question add sgml a 3 e To add an HTML tag a with attributes 8 to the string generated from the subtree rooted in the node in question add between sgml a between sgml a 3 23 2 Example e E Mere mi du This is lt A HREF http www cogentex com gt CoGenTex lt A gt VIG eee ae sap ase eo et gs eg ee eens DSYNTS BE1 I THIS2 number sg II CoGenTex class proper noun article no art sgml A HREF http www cogentex com END 24 POINTS TO CONSIDER WHEN MODIFYING THE GRAMMAR 56 23 3 Notes e To output the results of these features the for
5. of play The following is a complete list of modal auxiliaries which have entries in the lexicon CAN1 MAY1 MUST1 and SHOULD1 These modal auxiliaries need not be given any feature list Other modal auxiliaries should be specified as class verb modal aux note the space in their feature list An example is shown below 15 CLAUSES AND SENTENCES 34 e The copula be is treated as the head of its clause with the subject as actant I and the adjective noun or prepositional phrase which is predicated of the subject as argument II e For information on the argument structure in passive voice see Sec tion 15 3 page 36 e For information on wh questions see Section 17 page 40 14 4 Shortcomings e The subjunctive is not handled 15 Clauses and Sentences 15 1 Description Clauses and sentences are constructed by giving arguments to verbs The arguments of the verb are labeled I II III and IV I always corresponds to what is usually called the subject and II to IV correspond to objects of decreasing proximity to the verb direct object indirect object additional complement The feature extrapo governs the realization of the basic sentence structure Currently two variants are supported extraposition of the subject with anticipatory it It bothers me that she is there and there insertion with existential type verbs There appeared three geese in the study Feature extrapo can have the following values Subject
6. page 52 Extraposition of sentential subject Section 15 page 34 Section 16 3 page 39 Feminine gender cix x a net RUE DEPARA CA AA Section 9 page 19 POP LI tit Aste A II Es Leer Section 23 page 55 Ia e aa eei Ead E APR Section 14 page 30 Gender oF NOUNS napa ls Section 9 page 19 Genitive case cece eee Section 9 page 19 Section 11 page 26 Grammatical case A LIS fe et da Section 9 page 19 Humana features as e a IM MER Section 9 page 19 HTML formatting fue sae ie A A Section 23 page 55 Hyperlink usais Saori oa eee Gare satis Section 23 page 55 26 INDEX 60 IMPEraliVe serrat tinte orar aa Section 14 page 30 Indefinite determiner 00 cece cece eens Section 9 page 19 Indirect object Section 15 page 34 Section 15 3 page 36 A O dc a epa Section 18 3 page 45 it anticipatory subject Section 15 page 34 Section 16 3 page 39 TAA CS esca Crate a aedes rac ee dd Sey tes Section 23 page 55 Lec ata trip Es ee E TE eb ee eR dA Section 5 page 13 Hower CASO iss tan tur dubie doctr Rata BAM n en e o pad Section 21 page 51 Maseuline sender cestoagtnaansalos dad ite oce o Section 9 page 19 many determiner A SR ERES CN Rp SEES Section 10 1 page 22 may modal auxiliary rial Section 14 3 page 33 Modal abalar rd returns between a Section 14 3 page 33 more determinada Section 10 1 page 22 more than Construction to rc Section 18 3 page 45 most Cd obermIBeblo a
7. 1 class personal_pronoun person ist number sg IT eat class verb mood pres part I John class proper noun II bean class common noun number pl article no art 16 EMBEDDED CLAUSES 38 END Second example OUTPUT I told Mary that John eats beans END DSYNTS tell class verb tense past morpheme told inflection inv I 1 class personal_pronoun person ist number sg II Mary class proper_noun III eat class verb I John class proper noun II bean class common noun number pl article no art ATTR THAT3 END Third example OUTPUT I told Mary that John eats beans END DSYNTS BE1 class verb mood cond extrapo I I see class verb extrapo mood inf to extrapo I John class proper noun ref J1 II John class proper noun ref J1 gender masc 16 EMBEDDED CLAUSES 39 II horrible class adjective END Fourth example OUTPUT It bothers me that John can not see himself END DSYNTS bother class verb extrapo i I see class verb polarity neg I John class proper noun ref J1 II John class proper noun ref J1 gender masc ATTR THAT3 ATTR CAN1 II Mary class proper noun END 16 3 Notes e Sentential subjects are treated in the same manner as sentential objects just make them the first actant I If the sentential subject has MOOD TO INF and it has a first actant then a for is automatically inserte
8. Section 14 page 30 that complementizer ti AS Section 16 page 37 that demonstrative determiner sees Section 9 page 19 ie determine au ii oda Section 9 page 19 LOPES insertion ac Section 15 page 34 this demonstrative determiner ssesssess Section 9 page 19 UPDEr CASO Separar S26 fae Pg eti TOR ee ede ees EE Section 21 page 51 A a RR PU uds NT Section 14 page 30 VOICE ea set awl p Voti ce eu A a Gere eh an Section 14 page 30 Ah questionis iere dece E ES EXE DUREE wx eem iovis Section 17 page 40 ANS e scuta Ec es Ed a c Cutie Mesi Section 17 page 40 which relative pronoun esses Section 20 page 48 Word A Ee e e d eis Section 15 page 34 Section 18 page 42
9. must be Geese are in the garden clause or NP y That she is here bothers me There are geese in the garden NP J It bothers me that she is there Clase 15 CLAUSES AND SENTENCES 15 2 Examples OUTPUT John loves Mary END DSYNTS love class verb tense pres inflection reg I John class proper noun II Mary class proper noun END Second example OUTPUT John tells Mary a story END DSYNTS tell class verb tense pres inflection reg I John class proper_noun III Mary class proper noun II story class common noun article def END Third example OUTPUT Have there not been firefighters available in this city END 35 15 CLAUSES AND SENTENCES 36 DSYNTS BE1 class verb extrapo there polarity neg question taxis perf I firefighter class common_noun number pl article no art ATTR IN1 II city class common noun article dem prox II available class adjective END For examples involving extraposition of sentential subjects see Section 16 2 page 37 15 3 Notes e The notion of subject is purely syntactic not semantic Thus in a passive sentence such as John was killed by the car John is the syntactic subject and hence gets the arc label I e An agent in a passive clause is not a syntactic actant To generate John was killed by the car the by the car must be specified as an adverbial clause see Section 18 pag
10. relation to the verb 6 3 Notes In the DSynt Grammar the ordering of the rules is relevant for a given transformation RealPro applies the first rule whose pattern matches Con sequently more restrictive rules should be specified before more general ones 6 4 Shortcomings There are no known shortcomings 7 SSynt Grammar 7 1 Description The surface syntactic grammar SSynt Grammar contains rules to trans form a SSyntS into a linearized deep morphological structure DMorphS To facilitate the implementation in RealPro the SSynt Grammar has been divided in two grammars one specifying how to linearize a governor and its 7 SSYNT GRAMMAR 16 dependents and the other specifying how to linearize the dependents them selves These grammars are presented in next two sections Further details on these grammars and their formalisms are given in the RealPro Resource Specification Reference Manual 7 2 SSynt Grammar for Governor Dependents Lineariza tion 7 2 1 Description The SSynt Grammar for linearizing a governor and its dependents SSynt Grammar Governor Dependent specifies the ordering between a governor and its syntactic dependents 7 2 2 Example A AAA O AR SO O A A A A programmer Y worked X Lb SAA A O po AA COD DE edere SSYNT RULE X predicative Y Y number n person p X class verb lt gt 4 X lt X CX class verb number n person p This rule states
11. relative clause The value of the feature must be the same it can be any arbitrary string 20 RELATIVE CLAUSES 49 20 2 Examples Example 1 OUTPUT I saw the man who was drinking a Martini END DSYNTS SEE class verb tense past morpheme saw inflection inv I I class personal pronoun number sg person ist II man class common noun article def gender masc ref r1 ATTR drink class verb tense past aspect cont I man class common noun article def gender masc ref r1 II Martini class common noun article indef END Second Example OUTPUT I saw the blokes who were drinking Martinis END DSYNTS SEE class verb tense past morpheme saw inflection inv I I class personal pronoun number sg person ist II bloke class common noun article def number pl gender masc ref r ATTR drink class verb tense past aspect cont 20 RELATIVE CLAUSES 90 I bloke class common noun gender masc article def ref rl number pl pro pro II Martini class common noun article no art number pl END 20 3 Notes e You must indicate all relevant features on both nodes in the tree For example if we omitted the feature NUMBER PL in the lower bloke node in Example 2 REALPRO would generate saw the blokes who was drinking Martinis e The lexeme given to the lower of the two co referential nodes is actually irrelevant We have used the same lexeme as on the higher of the two nodes s
12. were written than Mary had thought that Mona could file with out reading but More articles were written than Mary regretted that Mona had read The DSyntS must be carefully constructed to gener ate the correct gappings Note that VP ellipsis is not currently handled Lyn wrote more papers than Steve did e No punctuation is added in any of the three positions To add commas around i e before or after an adverbial phrase use between punct comma See Section 22 1 page 52 for details e The mapping from information status theme rheme and so on to word order and other linguistic means of expression is a complex task which is not part of the tasks of REALPRO Future releases will have a separate module which performs this task 18 4 Shortcomings e For feature position values sent initial and sent final are not yet implemented e Positioning of adverbial phrases in English is a notoriously difficult problem and the current treatment is only a beginning In particular 19 COORDINATION AT it is possible in English to add adverbial phrases between auxiliaries John has often been admired This is not currently supported e In more than constructions the complex syntactic dependencies be tween the more and possible gaps in the than clause are not modeled in REALPRO More articles were written than Mary had thought that John could file without reading but More articles were written than Mary regretted that John had re
13. word END Second example OUTPUT The several ducks END DSYNTS duck class common noun article def ATTR SEVERAL END Third example OUTPUT All the ducks END DSYNTS duck class common noun article def ATTR ALL END 25 11 THE POSSESSIVE CONSTRUCTION 26 10 3 Notes e Quantifiers do not remove any articles or other determiners from the head noun in order to allow many a woman or all the women How ever REALPRO would also generate many the woman and all some women it is up to the specification of the input DSyntS to avoid such constructions e Numerals remove any indefinite articles from the head noun but allow for the specification of a definite article the four ducks e When specifying a quantifier not in the lexicon specify class quantificator note the non standard name for the class When specifying a numeral not in the lexicon use class numeral e To add more than to a numeral add MORE2 as an ATTR to the noun and then add the numeral as a II to the MORE2 See the first example above 10 4 Shortcomings The complex interaction among determiners fewer than five of those many nice young linguists has not been fully implemented In particular the appearance of ofin certain combinations most of the 56 men is not handled automatically by the grammar It is not clear how to force it either 11 The Possessive Construction 11 1 Description Nominal construction
14. ASS AS O SS this is a test 22 PUNCTUATION 52 DSYNTS BE1 caps none I THIS2 number sg II TEST2 article indef END 21 3 Notes None 21 4 Shortcomings There are no known shortcomings 22 Punctuation 22 1 Description By default the system generates a sentence with a final period unless fea ture question is used Section 14 page 30 in which case the sentence ends with a question mark This behavior can be overridden by adding the following features to the root node of the DSyntS representing the sentence e punct no_dot eliminates the sentence final period e g for titles e end punct question mark ends the sentence with a question mark P e end punct exclamation point ends the sentence with an exclama tion mark Cpi e end punct semicolon ends the sentence with a semicolon 22 PUNCTUATION 53 e end punct ellipsis dots ends the sentence with suspension points Cis Furthermore a bullet can be placed in front of a sentence e begin punct bullet begins the sentence with a bullet In addition parentheses brackets or quotes can be placed around an output sentence e between punct parenthesis puts parentheses around the sentence e between punct square bracket puts square brackets around the sentence X ei b e between punct double quote puts double quotes around the sentence e between punct sing
15. EVE E ER Educ Section 22 page 52 COMMA ds Section 22 page 52 COMMON NOWH e e reso ce ve Euer quer RU I Eo a Section 9 page 19 Complementizer tn dd ns Section 16 page 37 Compound Noun Cissa sas bs ir ees Section 9 3 page 21 Compound tense cce coe RUE A Section 14 page 30 Comtrol veria te Sone 2 dd ad Seid DR Section 16 page 37 Coordination dd Section 19 page 47 26 INDEX 99 CODUla srete ea eane A ted Bee ene ea MR ITI Section 14 3 page 33 D tivye cas dos dee A iba werk cd Section 9 page 19 Dativers hiit uaa e ple dd Sted Section 15 3 page 36 Definite determiner a ES eee eee ok Section 9 page 19 Definite noun phrase cdi epic er P eb EP RE Section 10 page 22 Demonstrative determiner 0 cece eee eee eee Section 9 page 19 Dependency it ee pitan eee ds gar ee Section 2 page 7 Descriptive relative clause 02 cece eee eee Section 20 page 48 Determiner sues Section 9 page 19 Section 10 page 22 ATEO JCC A p C ROME Sd RN LA ado Section 15 page 34 Double object construction eeeeeeeeess Section 15 3 page 36 D al pender educa me pee IR M iD dx Section 9 page 19 Echo QUESTION tarta Section 17 page 40 ESE NS uci Fs Section 16 page 37 Embedded clausii Unidad A pa A Section 16 page 37 Embedded wh question 0 000 cece eee eee ee Section 17 page 40 Example directories tl Gade a a ed Section 1 page 7 Exclamation Mark casadas perda tiza els Section 22
16. REALPRO General English Grammar User Manual August 30 1998 CoGenTex Inc 840 Hanshaw Road Suite 11 Ithaca NY 14850 1589 Tel 607 266 0363 Fax 607 266 0364 realpro cogentex com CONTENTS Contents 1 About this Document 2 Background Syntactic Dependency 3 DSyntS Format 3 1 3 2 3 3 3 4 4 Nodes 4 1 4 2 4 3 4 4 Description abs DE sechs eue et wem CE et Examples gos dunk es E op dE wee dote dod DD veia NOTES ei quac iade o ropa qute etta quise af o doe DIOPIODIBITIBS ua ER ERR Red Exodo S 5 The Lexicon 6 1 6 2 6 3 6 4 10 10 10 12 12 13 13 13 13 14 14 CONTENTS 7 SSynt Grammar Tal Description xxm AAA AAA 7 2 SSynt Grammar for Governor Dependents Linearization 7 2 1 Description dt le gets Mea e Hee E22 e o A A ae ee qur dior ER AAA A A A id 7 2 4 Shortcomings 1a xd e eee ee ak eros 7 3 SSynt Grammar for Dependents Linearization Goel Deseriphion e 3 aue ex wh RUD AAA 4 922 Examples ui ox e RERO Re LAE EORR add cANORES A duc x eese sx 7 3 4 Shortcomings Lg do Xo e A DEO XO bo Nod x 8 Defaults Bak o DOSOPLDETOTS e oodd a V op utet ee e Pie ek 5 Example Ss uude RIS EIE ao a a d urere A EL derat aa ed dS RU RE S RE SPON E A 8 4 Shortcomings hs E es ht RH Rue Pss Ba 9 Nouns 9 1 Description 4 io bob ku eC E E NO dedans 920 cath Ies acid guo poor ot do eco Se ce Gey d OE dem 2 Oro NOUR AS AUR dro rai a E dte 20 osos dece Efe DA ShoPUCOIngs a a
17. accusative cases does not exist overtly Instead we use the term objective case to cover both case obj 10 DETERMINERS 22 e Features for case are added by the grammar as needed The only times case features should be specified in the input to REALPRO are case gen in the possessive construction see Section 11 page 26 not yet implemented and case obj for the AcI construction see Section 16 page 37 e For other determiners numerals demonstratives those four tiaras see Section 10 page 22 e For the possessive construction John s tiara see Section 11 page 26 e For compound nouns diamond tiara relate the two nouns using ATTR You can also specify a single noun node diamond tiara e To append material after nouns use the APPEND relation For example in my son Desmond Desmond depends on son by an APPEND arc This arc label can also be used to add parentheses see Section 22 for details on using parentheses e Feature human is not used in REALPRO for English instead the rele vant distinctions can be made using gender dual or gender neut 9 4 Shortcomings e Specification gender dual should yield pronominal choices such as his or her but is not currently implemented e case gen is not currently implemented e Proper nouns are currently not inflected at all 10 Determiners 10 1 Description As mentioned in Section 9 page 19 at the deep syntactic level the definite indefini
18. ad 19 Coordination 19 1 Description To coordinate nodes X and Z with a conjunction Y use the relation COORD between X and Y and the relation II between Y and Z 19 2 Example ie E John laughed but Mary smacked the butler and the maid SR oec o Re see ae eee ee a ene a po qn DSYNTS laugh class verb tense past I John class proper_noun COORD BUT II smack class verb tense past I Mary class proper_noun II butler class common_noun article def COORD AND2 II maid class common_noun article def 20 RELATIVE CLAUSES 48 END 193 Notes e If two verbs are coordinated with shared post verbal arguments and adjuncts John bought and ate beans all year in Paris then these arguments and adjuncts should be dependents of the second lower verb and not of the first higher verb e REALPRO will determine that a constituent coordinated with and has plural agreement behavior 19 4 Shortcomings There are no known shortcomings 20 Relative Clauses 20 1 Description To form a relative clause add a complete well formed finite clause as a dependent to a nominal node e Use the ATTR arc to obtain a restrictive relative clause 1 saw the man who was drinking a Martini use the DESC ATTR arc to obtain a descriptive relative clause 1 saw the man who was drinking a Martini e There must be a feature ref specified for the hosting noun and the first or second argument of the
19. also possible to specify the preposition ON1 in the input specification to REALPRO The main advantage of specifying the preposition in the lexicon is that in this case the preposition is not meaning bearing he worked on the paper with the relevant non spatial meaning does not contrast with for instance he worked under the paper which only has a spatial meaning Ideally the input to REALPRO should only contain meaning bearing nodes e In specifying a lexical entry from the lexicon in a DSyntS it is not necessary ti use upper case We do so in this document for clarity e Lexical entries need not finish with an integer e g CAN1 This is only necessary if there potentially are different lexemes with the same label 5 4 Shortcomings There are no known shortcomings 6 DSynt Grammar 6 1 Description The deep syntactic grammar DSynt Grammar contains the rules used to transform a DSyntS into a SSyntS Surface Syntactic Structure Further details of the formalism used to specify the DSynt Grammar are described in the RealPro Resource Specification Reference Manual 7 SSYNT GRAMMAR 15 6 2 Example A SS GR c c EC A programmer Y worked X E ME C en CE C rg RD ICE DSYNT RULE X I Y CX E class verb lt gt X predicative Y 1 This rule states that the first I dependent Y of the verb X i e the subject should be carried over to the SSyntS as a dependent in the PREDICATIVE
20. ar immediately pre verbal John often eats beans More precisely the position is that immediately preceding the main verb of the clause whether finite or not except in the case of copular or existential be in which case the default position is immediately post verbal D sir e is often in the garden The position of the adverbial phrase can be controled through the use of the feature position marked on the head of the adverbial phrase It can have the following values 18 ADJUNCTS TO A CLAUSE 43 Often John eats beans John often eats beans sent final John eats beans often for clauses and prepositional phrases Note feature values sent initial and sent final are not yet implemented Please use the features starting_point and rheme described below In addition there are features that refer to the information status theme rheme and so on of phrases Currently there are two options e starting_point with value positions the adverbial phrase in sentence initial position Often John eats beans Thus starting point is a synonym for position sent initial e rheme with value positions the adverbial phrase in sentence final position John eats beans often Thus rheme is a synonym for position sent final 18 2 Examples First example _ John often eats beans _
21. atches Consequently more restrictive rules should be specified before more general ones 7 3 4 Shortcomings To be supported by the current version of RealPro a rule regarding dependent dependent linearization must be between one governor and two and only two dependents 8 Defaults 8 1 Description Default features are added to lexemes when the DSyntS is read These features are specified in the file defaults dat in the directory LKB 8 2 Example This is the standard default file supplied with REALPRO DEFAULT verb tense pres mood ind DEFAULT common_noun noun number sg person 3rd gender neut article indef DEFAULT proper noun noun number sg person 3rd gender neut article no art 8 3 Notes e The user can change the file defaults dat freely the required syntax is self explanatory Defaults can be added to any lexical class 9 NOUNS 19 e The term standard defaults refers to the defaults set in the file defaults dat with the delivered system e In this document the standard defaults are listed in the sections for the relevant lexical classes 8 4 Shortcomings There are no known shortcomings 9 Nouns 9 1 Description There are two subtypes of nouns common nouns and proper nouns Nouns have four types of features for number for gender for case and for determiners e Feature number can have the following values E hem pi beans e Feature gender can have the following values
22. d for the first actant e To extrapose a sentential subject and replace it by an expletive it use extrapo i on the main verb not on the verb of the sentential subject 17 WH QUESTIONS 40 See 15 page 34 e You must indicate the verb form of the embedded clause using the MOOD feature e Complementizers and subordinating conjunctions such as that should be added as ATTR dependents to the embedded main verb The lexical entry for the complementizer that in the lexicon is THAT3 See the second example above 16 4 Shortcomings e The example given above is in fact an Acl or raising to object or ECM construction meaning that the embedded subject is in accusative case I saw him eating beans Currently the case must be marked manually in the input specification e Raising verbs raising to subject such as to seem are not currently handled correctly Instead they can be treated like control verbs 17 Wh Questions 17 1 Description Wh questions are questions that involve at least one wh word To generate such a sentence use pro wh on the argument or adjunct that is to be a wh word 17 2 Examples First example OUTPUT Who likes John END 17 WH QUESTIONS 41 DSYNTS like class verb I Manfred class proper noun gender masc pro wh II John class proper noun END Second example OUTPUT The authorities are wondering who gave books to whom END DSYNTS wonder
23. e OUTPUT This is a test END DSYNTS BE1 I THIS2 number sg II TEST2 article indef 3 3 Notes e Indentation and line breaking is not relevant We follow this formatting convention only in order to make the tree like structure evident e Comments can be added before or after the DSyntS specification i e before the keyword DSYNTS or after the keyword END Comments can consist of any strings except the keyword DSYNTS e For testing purposes the target surface form can be declared before the specification by surrounding it by the keywords OUTPUT and END 4 NODES 10 This specification is used during regression testing to automatically check for discrepancies between the target surface form and the realized surface form 3 4 Shortcomings There are no known shortcomings 4 Nodes 4 1 Description A node is specified as follows e A specification of a lexeme followed by e a list of features A lexeme is e either a lexeme not in the lexicon or e a lexeme in the lexicon For lexemes in the lexicon see Section 5 page 13 A lexeme not in the lexicon is specified by its root form uninflected Lexemes with regular morphology and regular syntactic behavior typically are not included in the lexicon The capitalization of the input version is carried over to the output Several independent words can be combined by underscor
24. e 42 e As a default the third actant marked by III is realized as an indirect object rather than as a prepositional object see the second example above unless the third actant is marked rheme in which case it is realized by default as a prepositional object with to see the third example above These defaults can be overridden by entries in the lexicon see Section 5 page 13 e Verbs can be specified in the lexicon for having a strongly governed preposition introducing one or more of their actants They discriminate against foreigners If there is no entry in the lexicon the preposition must be added in the DSyntS 16 EMBEDDED CLAUSES 3T 15 4 Shortcomings e Relation IV is not presently supported e For there insertion clauses if the number of the DSyntS subject i e the I argument is not marked in the input DSyntS it will default to singular no matter what numerals or quantifiers have been specified 16 Embedded Clauses 16 1 Description Embedded clauses are formed simply by adding the embedded clause as an argument to the matrix verb Depending on whether or not the matrix verb also has a nominal object this will be as second or third argument using II or III respectively The verb form of the embedded clause s main verb must be explicitly marked on the verb 16 2 Examples First example OUTPUT I saw John eating beans END DSYNTS SEE class verb tense past morpheme saw inflection inv I
25. er aed Section 16 page 37 D rncbudtQ A ia ed Section 22 page 52 o A A A A RC O Section 10 page 22 Questio ce F s une wed oe bee ii ras ne Heese a XpiS PTS Section 14 page 30 Question Moor Section 22 page 52 A Section 22 page 52 PRAMS TT VER id sa cedo A oc E NE Section 16 page 37 Reduced relative clause ooooooooooooooooo Section 20 3 page 50 Reflexive pronoun 0 00 cece eee eee eee teens Section 12 page 27 Relative clause eg remped ese ERAS et ve rid Section 20 page 48 Relative pronoun ii Section 20 page 48 Restrictive relative clause 0 cece eee eee eee Section 20 page 48 Rhemes ao be che Wa oe s Section 15 page 34 Section 18 page 42 s Anglo Saxon genitive 0 cece cece eee eee Section 11 page 26 A e s eon ewe eee EE ASSES DRE e Section 22 page 52 Sentential subject espiadas do Section 16 3 page 39 should modal auxiliary ws darse ro Rer Tits Section 14 3 page 33 Singular number a sada ita Ree Da Section 9 page 19 some determiner Section 9 3 page 21 Section 10 1 page 22 Subcategorization frame ss oo crm T Ra ees Section 15 page 34 DUDTE D GELVE uci AA Ceca ied E ad Section 14 page 30 Subordinate clauses 24 cei Pee ewe teed Section 16 page 37 26 INDEX 62 Subordinating conjunction Section 18 3 page 45 Section 16 page 37 Syntactic dependency ooocccocoorcccrno rec serena Section 2 page 7 gr rr Section 14 page 30 O rem A An RP eade
26. es which are converted to spaces in the output A list of features is specified by e An open bracket followed by 4 NODES 11 e a possibly empty list of feature value pairs of the form feature value separated by spaces followed by e A close bracket 1 Features are optional if defaults are provided see Section 8 page 18 except that a lexeme which is not in the lexicon must have the class feature For open class words feature class can have the following values adjective small disastrous adverb really fast common_noun table map proper_noun John Poona Socks verb to play to indulge symbol For closed class words feature class can have the values shown below Note that all of these lexemes are in fact in the lexicon and should be specified in that manner see Section 5 page 13 this table is given for informative purposes only article coordinative com demoustrative pronoun numeral particle partitive pronoun preposition all quantificator E subordinativc con 4 NODES 12 4 2 Examples First example A A a es RS OS ia il Mesmerizingly ia E CR dm DSYNTS mesmerizingly class adverb END Second example Jg CR CN GERE GR E gueule erc amp FuN aNd GaMeS amp A Se See oer CIE DSYNTS gFuN_aNd_GaMeSg class proper_noun END 4 3 Notes e The spacing does not matter for the lexeme feature list combination e The ordering of the feature val
27. f New York Press 1988 26 INDEX 98 26 Index a determiner ce be beng ee 8 cect ats Sensit hie ake Section 9 page 19 Accusative CASO io era Hing VER RS Section 9 page 19 Ac construction A duis ates Saw do M te Section 16 page 37 a cas doa rans ead SN ila c E lis Ta eget Ms Section 14 page 30 Adjunct clause to a clause 00 0 cee eee eee eee Section 18 page 42 Adjunct clause to a noun 6 eee eee eee Section 20 page 48 Adjunct to arcas rada Section 18 page 42 PAN OTS rra gas bad sd asse TE bege Section 18 page 42 Adverbial doe S es Ro ceto etico EO noc Section 18 page 42 Adverbial phrase 22422 f queues ew rbd de Section 18 page 42 Agent in a passive clause i ven Epp paris pra Rt Section 15 page 34 all determiner uires i estie EN Section 10 1 page 22 DIL e code pe s ote eoo AN abt UN dnd Section 19 page 47 any determiner std Section 10 1 page 22 Anticipatory subject uto ccce ce vxo eee wo bis Section 16 3 page 39 Pao ET Section 9 page 19 Baro DIE er dd oa cio Section 9 3 page 21 ATAN PR sat E af NE Section 14 3 page 33 Bindinetheoly sia Si detis eas Rated A RR A Section 12 page 27 ONG TACOS qu as E Adi Section 23 page 55 DECESO RR LE od LUE Section 22 page 52 can modal auxiliary ass Sato e De detis Section 14 3 page 33 CapitallzatlOB sissies ecaa Ex EO uice er d D a edili Section 21 page 51 Cardinal number vec oe ed cn leot a ated aed ides Section 10 page 22 Colon doses oe ter B Ro E CE
28. ggest have not yet been implemented through features They can of course be obtained simply by specifying them in the input 14 Verbs 14 1 Description This is an exhaustive list of verbal features 14 VERBS 31 e Feature tense can have the following values John likes Mary John liked Mary John will like Mary EN Feature tense is not meaningful in conjunction with a non finite mood see below REALPRO will usually ignore the value of the tense fea ture e Feature voice can have the following values sample pass Jom isika For information on the argument structure in passive voice see Sec tion 15 3 page 36 w e Feature aspect can have the following values xample Default simple John eats beans cont John is eating beans e Feature taxis can have the following values Value Dci nil John likes Mary perf John has liked Mary e Feature mood can have the following values les 14 VERBS 32 John likes Mary John would like Mary mp Ca Mary S For John to like Mary would be a problem pres part John liking Mary is a problem Given the book Mary disappeared pee Es John Tike Mary E E The subjunctive lest John cause trouble in the present if John were mistaken in the past is currently not supported However the present subjunctive is always morphologically identical to the bare infinitive in English The combination of imperative and ques
29. ince this clarifies the situation best e To obtain a reduced relative clause a passive relative clause in which the relative pronoun and the passive auxiliary are omitted such as the blokes attacked by Mary specify mood past part on the verb and do not include the actant that is omitted For example OUTPUT The blokes attacked by Mary END DSYNTS II bloke class common noun article def number pl gender masc ref r1 ATTR attack class verb mood past part ATTR BY1 II Mary class proper noun article no art 21 CAPITALIZATION 51 END 20 4 Shortcomings e A serious bug in the current version is that the lower node cannot be labeled with a lexeme in the lexicon If the node is labeled with a lexeme in the lexicon then the relative pronoun will not be generated This will be fixed in the next release e The co referential noun must be an immediate dependent of the verb of the relative clause there is no pied piping to obtain the man whose tiara was stolen or a situation up with which I will not put 21 Capitalization 21 1 Description e Use caps none on a node to keep the word generated from that node from being capitalized for example if it appears in sentence initial position e Use caps words to capitalize all words generated from the subtree rooted in the annotated node e Use caps word to capitalize just the word generated from the anno tated node 21 2 Example JE SH
30. le quotes puts single quotes around the sentence These features can also be used at other nodes in a DSyntS The parentheses brackets or quotes are then placed around the text string generated by the subtree dominated by the annotated node Comma punctuation within a sentence is handled by the grammar Addi tional commas can be added using the following features e between punct comma puts commas around the string that is gen erated from the subtree rooted in the annotated node e leftmost punct comma puts a single comma after the word imme diately preceding the text string generated by the subtree dominated by the annotated node e rightmost punct comma puts a single comma after the last word of the text string generated by the subtree dominated by the annotated node The same commands with comma replaced by dash can be used to generate dashes 22 PUNCTUATION 94 22 2 Example aaa SRU CU CR John loves Mary is a e E ec EEN DSYNTS love class verb tense pres inflection reg between punct parenthesis I John class proper_noun II Mary class proper noun END 22 3 Notes e The system automatically does point absorption If several point punctuation marks period semicolon colon dash comma coincide in the same location the point with the highest precedence is chosen The priority hierarchy is as follows period precedes semicolon precedes
31. matting must be set to HTML e Surround the string representing the attributes of an HTML tag with double quotes if it contains a colon eg syml A HREF http www cogentex com or if it contains spaces 23 4 Shortcomings There are no known shortcomings 24 Points to Consider When Modifying the Grammar There are some dependencies regarding the knowledge encoded in the lexicon and the various grammars these linguistic resources share similar labels for the features lexemes syntactic relations etc In the current version of RealPro it is the task of the developer to ensure that any modification s he does to one linguistic resource is consistent with the information found in the other linguistic resources Here is a list of some typical items to verify when the grammar is modified e Verify that the deep syntactic relations used in a DSyntS are covered in the DSynt Grammar e Verify that the deep syntactic relations introduced in the government pattern of a lexical entry are covered in the DSynt Grammar e Verify that the surface syntactic relations introduced in a DSynt Grammar rule are covered in the SSynt Grammars 25 OTHER DOCUMENTS oT 25 Other Documents Here is a list of documents currently available or that will be available soon e RealPro Resource Specification Reference Manual Jan 1997 e RealPro C API Reference Manual Jan 1997 e Dependency Syntax Theory and Practice I Mel uk State University o
32. os 20 Relative Clauses 37 3T 3T 39 40 40 40 40 41 42 42 42 43 45 46 47 AT AT 48 48 48 CONTENTS 20 1 Dese DEO y loe Rote Nen Vi opa 20 2 Examples seu Gig ow ck Ere A deb A oe A RCM ZEN n I CM 20 4 Shortcomings pcs tale A oras i eut bx tdg 21 Capitalization 21 1 D s ertiptioh sans aux Cooks Ero ESSE WEE OE S Dak AGAIN Gs ig otra erede wee RE RERUM E EC a E 21 5 NOTES de zx Te epe o Emp estu e A ee wo E 21 2 ShoBLCODILABS aua di ie uo E ese bee RD AR Be ied 22 Punctuation 22 1 Description uo A e BECA deed abus Cs DEDO E AC soe us aedes io dete e oe SES e CoD tse a acca dM 22D AN OEA Eodem b deo ne odeur weg tei a ee tie oe a 22 2 OSTOFDCORILES S ose ere gie ee a d uo do 23 HTML Annotations Zoe Description ALAS oe 29 2 A eoo s es La era n dus o dl Le god 29 0 NOE e leds a xo us 29 DOT CONMNESO seringe owie e Magui a GENS acids 24 Points to Consider When Modifying the Grammar 25 Other Documents 26 Index 51 51 51 52 52 52 52 94 54 54 55 95 95 56 56 56 57 58 1 ABOUT THIS DOCUMENT i 1 About this Document This document describes the linguistic resources which make up REALPRO s general English grammar More information about the formalisms used to specify the linguistic resources can be found in the RealPro Resource Speci fication Reference Manual for more information about the C application programmer s interface see the RealPro C API Reference Man
33. os aoa i oH xut HO RUE XE 10 Determiners 10 1 Description E ue RAS oe Rer oen PUE ee SERRE ES 10 2 Examples mesna ed cose ek E E s O a ee Bd 15 15 16 16 16 16 17 17 17 17 18 18 18 18 18 18 19 19 19 20 21 22 CONTENTS 10 3 Notes 10 4 Shortcomings 11 The Possessive Construction 11 1 Description 11 2 Example 11 8 Notes 11 4 Shortcomings 12 Pronouns 12 1 Description 12 2 Examples 12 3 Notes 12 4 Shortcomings 13 Adjectives 13 1 Description 13 2 Example 13 3 Notes 13 4 Shortcomings 14 Verbs 14 1 Description 14 2 Example 14 8 Notes 14 4 Shortcomings 15 Clauses and Sentences 15 1 Description 26 26 26 21 27 27 27 28 29 29 29 29 30 30 30 30 30 33 33 34 34 CONTENTS 15 2 Examples loa de ae TSS Notes E a E 15 4 Shortcomings cu a 09 pda ga 16 Embedded Clauses 16 1 Description o sne s Sue Re uox Ee 16 2 Examples yb utu E re i re Ib NDS AAA 16 4 Shortcomings AR 17 Wh Questions 17 1 Description epi a 17 2 Examples 4 03 a SE TeS INOUOSL os odd Ro O ee 17 4 Shortcomings 225 1 2 id mE Stew 18 Adjuncts to a Clause 18 1 Descriptor el Ros AA bee 18 2 Examples a a ta 15 9 Nba 4 4 3 acea faeta srt ep ets 18 4 Shortcomings ps kk Rp mun 19 Coordination 19 1 Description poene d oe dope 19 2 Example ii ox oe 3e xo dE epe 10 9 INGLES x v is eR bed Rein Roe GE ES 19 4 Shortcomings uci os ence Exo e
34. rb I John class proper noun ref J1 II John class proper noun ref J1 gender masc END Third example OUTPUT The psychiatrist revealed the patient to herself 15 ADJECTIVES 29 END DSYNTS reveal class verb tense past I psychiatrist class common noun article def II patient class common noun ref P1 article def III patient class common noun ref P1 article def gender fem END 12 3 Notes e If the second and third actant are co referential REALPRO always generates the third actant as a to phrase irrespective of the rheme feature with a reflexive pronoun e The grammar overrides any indications of article on a pronoun to be pronominalized as in the third example above 12 4 Shortcomings e While the category partitive pronoun exists in the lexicon it does not have any special syntactic behavior 13 Adjectives 13 1 Description Adjectives can get attached to nouns in two manners e By the ATTR arc in which case they appear pre nominally the usual case 14 VERBS 30 e By the DESC ATTR arc in which case they appear post nominally set off by commas 13 2 Example E DEA A AAA A AAA A Two eggs small SEA os O O o ag eee DSYNTS egg class common noun article no art number pl ATTR TWO DESC ATTR small class adjective END 13 3 Notes ATTR and DESC ATTR arcs are repeatable 13 4 Shortcomings The comparative bigger and the superlative bi
35. s with a possessor such as John s tiara 11 2 Example Forthcoming 12 PRONOUNS 27 11 3 Notes This construction will be implemented soon 11 4 Shortcomings This construction is not yet implemented 12 Pronouns 12 1 Description Pronouns are generated by the morphological component based on features present on nodes There are basically two ways of specifying the use of a pronoun e Use a special lexical entry such as lt PRONOUN gt This lexical entry introduces special features and handles the absence of articles The entry still needs to specify number and person e Use a feature on a noun In this case the noun is realized as a pronoun The following pronominalizations are handled by the grammar and should not normally require annotation in the input DSyntS e heflexives are added by the grammar in cases in which their appearance is determined grammatically Specifically if the 1st and 2nd actant have the same value for the ref feature or the first and third or the second and third then the lower argument is replaced by a reflexive e Relative pronouns are added by the grammar See Section 20 page 48 for details Here is a list of the types of pronouns currently supported 12 PRONOUNS 28 pro pro lt POSSESSIVE PRONOUN gt pro poss lt REFLEXIVE PRONOUN gt pro rel 12 2 Examples First example Item missing Second example OUTPUT John sees himself END DSYNTS see class ve
36. tS is a deep syntactic representation meaning that only meaning bearing lexemes are represented and not function words This means that REALPRO does not perform the task of lexical choice the input to REALPRO must specify all meaning bearing lexemes Furthermore there is no non determinism in REALPRO since the rules are applied in the order in which they are defined without backtracking This means that the input to REALPRO fully determines the output but it represents it at a very abstract level which is well suited for interfacing with knowledge based applications 3 DSyntS Format 3 1 Description RealPro takes as input a DSyntS which can be specified either programmat ically or with an ASCII based formalism This section describes only the ASCII based formalism to represent a DSyntS The input is structured as follows e The keyword DSYNTS followed by e the specification of the deep syntactic structure DSyntS followed by e the keyword END The DSyntS is specified as follows e specification of a node followed optionally by e an open parenthesis an arbitrary non null number of dependent specifications followed by a close parenthesis A dependent is specified as follows e specification of a dependency arc 3 DSYNTS FORMAT 9 e followed by a specification of a node A dependency arc is specified simply by the arc label The node specification is explained in Section 4 page 10 3 2 Exampl
37. te and demonstrative articles are specified with the feature article 10 DETERMINERS 23 which can have one of the following values def indef dem prox dem dist or no art Other determiners should be added as ATTR dependents of the noun They fall into two categories quantifiers all many and so on and numerals one sir several and so on The difference between the two categories is that quantifiers are ordered before articles while numerals are ordered after articles all the boys any seven boys the one thing The following quantifiers can be found in the lexicon The table shows the number agreement that is forced on the head noun by the quantifier Entry in lexicon Number agreement The following numerals can be found in the lexicon The cardinal numerals can be referred to either by a spelled out lexeme or by an integer The table shows the number agreement that is forced on the head noun by the quantifier 10 DETERMINERS 24 Entry in lexicon Alternate integer representation Number agreement ZERO ONE TWO THREE FOUR FIVE SIX SEVEN EIGHT NINE TEN ELEVEN TWELVE SEVERAL CON DOR QN The digital version of the lexeme but not the full word version has a feature form which can take the following values weve E 10 2 Examples First example OUTPUT More than six ducks END DSYNTS duck class common noun 10 DETERMINERS ATTR MORE2 II 6 form
38. that a dependent Y standing in the PREDICATIVE relation to the verb X which governs it should precede this verb in the linearization this rule also ensures that the verb agrees in number and person with X 7 2 3 Notes In the SSynt Grammar Governor Dependent the ordering of the rules is relevant for a given transformation RealPro applies the first rule whose 7 SSYNT GRAMMAR 17 pattern matches Consequently more restrictive rules should be specified before more general ones 7 2 4 Shortcomings To be supported by the current version of RealPro a rule rule regarding governor dependent linearization must be between one governor and one and only one dependent 7 3 SSynt Grammar for Dependents Linearization 7 3 1 Description The SSynt Grammar for linearizing dependents SSynt Grammar Dependent Dependent specifies the ordering between the syntactic dependents of a given governor 7 3 2 Example V a en ire ma mu Will Z the programmer Y work X d a O A O SSYNT RULE X predicative Y X auxiliary Z CX class verb invert lt gt X lt Y I E This rule states that the AUXILIARY dependent Z of a verb X bearing the feature INVERT should precede the dependent Y standing in the PRED ICATIVE relation to X 8 DEFAULTS 18 7 3 3 Notes In the SSynt Grammar Dependent Dependent the ordering of the rules is relevant for a given transformation RealPro applies the first rule whose pattern m
39. tion is not supported Subject auxiliary inversion only happens with finite auxiliaries e Feature polarity can have the following values John likes Mary John does not like Mary e Feature question can have the following values Deal John likes Mary Does John like Mary This feature is used in the input DSyntS only to specify yes no ques tions For information on wh questions see Section 17 page 40 E E 14 VERBS 33 14 2 Example A A MM c ae John does not love Mary A A A E DSYNTS love class verb tense pres polarity neg inflection reg I John class proper noun II Mary class proper noun END 14 3 Notes e These verbal features can be combined in any way though the non finite moods infinitive with or without to and the present and past participles and the conditional mood do not have tense and do not choose past present or future Specifications of tense in such cases will be ignored e For the imperative mood imp the generator will not automatically remove the subject in order to allow for constructions such as You be on time Furthermore the exclamation mark is not generated auto matically either See Section 22 page 52 e Modal auxiliaries such as can and may are not recursive in English can may come and therefore they should be attached as an ATTR to the main verb Put differently in John can play tennis John and tennis are arguments I and II respectively
40. u etoo ds dico Section 10 1 page 22 must modal auxiliary sus ea qi t es Section 14 3 page 33 Negation quse ters 2 ee eee cer Section 14 page 30 Neuter gender comicidad gas Section 9 page 19 Nominative case cce eon AA Xx CO E WU AR A Gd Section 9 page 19 INOUE iurc nct CALME TEL ACE e Ford LAM di Section 9 page 19 Noun compound di cti e hon itp Section 9 3 page 21 Number of nouns o Eee eor cd Section 9 page 19 Objective CASE a we tues surto pes usage E LAT E ET SEEEN Section 9 page 19 A acte I eoa state Section 10 page 22 DE ccm Section 19 page 47 Patentheses ps oco exe ER RH PIER ER EV Section 22 page 52 PASSIVO VOICE used edad eda eed Glos e E ee Section 14 page 30 Past tendo ei beer Des E e dee ae E EPI EUER Section 14 page 30 Pertective aspect ceri eeu een ape aa ime rs Section 14 page 30 Pernod ascensu a Deme tc bed E unm Section 22 page 52 26 INDEX 61 Personal Pronoun 4422 4 eisai ere ciega ade X HS Section 12 page 27 Pied o P Section 20 4 page 51 Plural number tas essa ND ras re debes ir ee Section 9 page 19 Possessive construction 22s ad Section 11 page 26 Possessive pronoun apar eee ERA ERES UE deed Section 12 page 27 Prepositional phrase modifying a clause Section 18 page 42 Prepositional object as qe eed e eet Section 15 page 34 Present tenge EA A RD a Section 14 page 30 PRONOUN do a B P EIE e Section 12 page 27 Proper noun cates oe e DO ed a Section 9 page 19 PROS dd at a ir a
41. ual All of the examples used in this document can be found in the directory of examples called DsyntS which can be found in the top level directory of the REALPRO distribution This directory contains several subdirectories which group related examples for example Noun and Punctuation The files fo rthe examples have descriptive names This document will refer to these examples using UNIX style pathnames starting in the DsyntS direc tory so for example Punctuation double quotes dss will refer to file DsyntS Punctuation double quotes dss 2 Background Syntactic Dependency The input representation to REALPRO is a syntactic dependency represen tation It is called the Deep Syntactic Structure or DSyntS for short and is proposed in this form by I Mel uk in his Meaning Text Theory This representation has the following salient features e The DSyntS is a tree with labeled nodes and labeled arcs e The DSyntS is lexicalized meaning that the nodes are labeled with lexemes uninflected words from the target language e The DSyntS is a dependency representation and not a phrase structure representation there are no nonterminal nodes such as VPs and all nodes are labeled with lexemes e The DSyntS is a syntactic representation meaning that the arcs of the tree are labeled with syntactic relations such as SUBJECT rather than conceptual or semantic relations such as agent 3 DSYNTS FORMAT 8 e The DSyn
42. ue pairs does not matter in the list of features e By default the output is formatted as a sentence with an initial capi talization and a final period To avoid the final period add punct no dot to the root verb which eliminates the sentence final period See Sec tion 22 1 page 52 for details To avoid initial capitalization use caps none See Section 21 page 51 for details 5 THE LEXICON 13 4 4 Shortcomings There are no known shortcomings 5 The Lexicon 5 1 Description Words can be specified in the lexicon This obviates the need for specifying in the input DSyntS information about irregular morphology and irregular syn tax The details of the formalism used to specify lexical entries are described in the RealPro Resource Specification Reference Manual 5 2 Example The following example gives a specification for the lexeme WORK LEXEME WORK CATEGORY verb FEATURES GOV PATTERN DSYNT RULE WORK II X2 E lt gt WORK completivel ON1 ON1 prepositional X2 MORPHOLOGY LE work reg This lexical entry specifies government pattern introducing the prepositional lexeme ON1 for the second II dependent of WORK e g A programmer worked WORK on ON1 the project X2 6 DSYNT GRAMMAR 14 5 3 Notes e The field CATEGORY represents what is called in this document the fea ture class e Instead of using the lexical entry above for WORK it is of course
43. ys the best with all types of adverbial phrases For example a prepositional phrase is usually not placed pre verbally John has in Paris eaten brains However they do not appear to be ungrammatical in that position John has in the past eaten brains e For adverbial clauses the subordinating conjunction including if should be treated as an ATTR of the adverbial clause s main verb The main verb of the adjunct clause itself should be the ATTR of the main clause See the fourth example above The following subordinating conjunc tions are in the lexicon EVEN IF IF THAN1 THAT3 THENI 18 ADJUNCTS TO A CLAUSE 46 e When using a subordinating conjunction which is not in the lexicon use feature class subordinative conj note the non standard termi nology e In an if then construction the then clause is the main clause and the if clause the adjunct clause Use IF and THEN1 Note that the then is only possible if the if clause is preposed starting point on the main verb of the if clause This is not enforced by REALPRO e For a more than construction the clause containing more is the main clause and the than clause is the adjunct clause Use MORE1 if an adverb She worked more than he though he would and MORES if a quantifier More children arrived than Billie Jean had expected Note that the complex syntactic dependencies between the more and possible gaps in the than clause are not modeled in REALPRO More articles
Download Pdf Manuals
Related Search