From 31e0deb017a938bc91f49d8505104d97bc8af14f Mon Sep 17 00:00:00 2001 From: aarne Date: Tue, 21 Oct 2003 15:15:47 +0000 Subject: [PATCH] Rebuilding resource libraries. Rebuilding resource libraries. Working with resource interfaces. --- grammars/resource/abstract/ResAbs.gf | 23 +- grammars/resource/english/Morpho.gf | 2 +- grammars/resource/english/Types.gf | 2 +- grammars/resource/finnish/Morpho.gf | 654 ++++++++++++++++++ grammars/resource/finnish/ResFin.gf | 215 ++++++ grammars/resource/finnish/Syntax.gf | 974 +++++++++++++++++++++++++++ grammars/resource/finnish/TestFin.gf | 40 ++ grammars/resource/finnish/Types.gf | 123 ++++ src/GF/Canon/Look.hs | 3 + src/GF/Compile/CheckGrammar.hs | 6 +- src/GF/Compile/Extend.hs | 28 +- src/GF/Compile/Rename.hs | 14 +- src/GF/Data/Operations.hs | 8 + src/GF/Grammar/Lookup.hs | 2 + src/Today.hs | 2 +- 15 files changed, 2072 insertions(+), 24 deletions(-) create mode 100644 grammars/resource/finnish/Morpho.gf create mode 100644 grammars/resource/finnish/ResFin.gf create mode 100644 grammars/resource/finnish/Syntax.gf create mode 100644 grammars/resource/finnish/TestFin.gf create mode 100644 grammars/resource/finnish/Types.gf diff --git a/grammars/resource/abstract/ResAbs.gf b/grammars/resource/abstract/ResAbs.gf index 7bb4d8816..7828e51ac 100644 --- a/grammars/resource/abstract/ResAbs.gf +++ b/grammars/resource/abstract/ResAbs.gf @@ -5,22 +5,26 @@ -- Although concrete syntax differs a lot between different languages, -- many structures can be found that are common, on a certain level -- of abstraction. What we will present in the following is an abstract --- syntax that has been successfully defined for English, French, German, +-- syntax that has been successfully defined for English, Finnish, French, German, -- Italian, Russian, and Swedish. It has been applied to define language -- fragments on technical or near-to-technical domains: database queries, -- video recorder dialogue systems, software specifications, and a --- health-related phrase book. --- +-- health-related phrase book. Each new application helped to identify some +-- missing structures in the resource and suggested some additions, but the +-- number of them was usually small. +-- -- To use the resource in applications, you need the following -- $cat$ and $fun$ rules in $oper$ form, completed by taking the --- $lincat$ and $lin$ judgements of a particular language. There is --- a GF command for making this translation automatically. +-- $lincat$ and $lin$ judgements of a particular language. This is done +-- by using the $reuse$ module with the desired concrete syntax of +-- $ResAbs$ as argument. + --2 Categories -- -- The categories of this resource grammar are mostly 'standard' categories -- of linguistics. Their is no claim that they correspond to semantic categories --- definable in type theory: to define such correspondences it the business +-- definable in type theory: to define such correspondences is the business -- of applications grammars. -- -- Categories that may look special are $Adj2$, $Fun$, and $TV$. They are all @@ -102,13 +106,14 @@ cat Phr ; -- full phrase, e.g. "John walks.","Who walks?", "Wait for me!" Text ; -- sequence of phrases e.g. "One is odd. Therefore, two is even." + --2 Rules -- --- This set of rules is minimal, in the sense defining the simplest combinations --- of categories and of not having redundant rules. +-- This set of rules is minimal, in the sense of defining the simplest combinations +-- of categories and not having redundant rules. -- When the resource grammar is used as a library, it will often be useful to -- access it through an intermediate library that defines more rules as --- combinations of the ones below. +-- 'macros' for combinations of the ones below. --3 Nouns and noun phrases -- diff --git a/grammars/resource/english/Morpho.gf b/grammars/resource/english/Morpho.gf index b5fcaa0fa..bd6ac1ca1 100644 --- a/grammars/resource/english/Morpho.gf +++ b/grammars/resource/english/Morpho.gf @@ -5,7 +5,7 @@ -- This resource morphology contains definitions needed in the resource -- syntax. It moreover contains the most usual inflectional patterns. -- --- We use the parameter types and word classes defined in $types.Eng.gf$. +-- We use the parameter types and word classes defined in $Types.gf$. resource Morpho = Types ** open Prelude in { diff --git a/grammars/resource/english/Types.gf b/grammars/resource/english/Types.gf index dd6682882..ad96db43b 100644 --- a/grammars/resource/english/Types.gf +++ b/grammars/resource/english/Types.gf @@ -6,7 +6,7 @@ -- However, it only includes those parameters that are needed for -- analysing individual words: such parameters are defined in syntax modules. -- --- we use the language-independent prelude. +-- We use the language-independent prelude. resource Types = open Prelude in { diff --git a/grammars/resource/finnish/Morpho.gf b/grammars/resource/finnish/Morpho.gf new file mode 100644 index 000000000..6aca6be61 --- /dev/null +++ b/grammars/resource/finnish/Morpho.gf @@ -0,0 +1,654 @@ +--1 A Simple Finnish Resource Morphology +-- +-- Aarne Ranta 2002 +-- +-- This resource morphology contains definitions needed in the resource +-- syntax. It moreover contains the most usual inflectional patterns. +-- +-- We use the parameter types and word classes defined in $Types.gf$. + +resource Morpho = Types ** open (Predef = Predef), Prelude in { + +--2 Nouns +-- + +oper + +-- worst-case macro + + mkSubst : Str -> (_,_,_,_,_,_,_,_,_,_ : Str) -> CommonNoun = + \a,vesi,vede,vete,vetta,veteen,vetii,vesii,vesien,vesia,vesiin -> + {s = table { + NCase Sg Nom => vesi ; + NCase Sg Gen => vede + "n" ; + NCase Sg Part => vetta ; + NCase Sg Transl => vede + "ksi" ; + NCase Sg Ess => vete + ("n" + a) ; + NCase Sg Iness => vede + ("ss" + a) ; + NCase Sg Elat => vede + ("st" + a) ; + NCase Sg Illat => veteen ; + NCase Sg Adess => vede + ("ll" + a) ; + NCase Sg Ablat => vede + ("lt" + a) ; + NCase Sg Allat => vede + "lle" ; + + NCase Pl Nom => vede + "t" ; + NCase Pl Gen => vesien ; + NCase Pl Part => vesia ; + NCase Pl Transl => vesii + "ksi" ; + NCase Pl Ess => vetii + ("n" + a) ; + NCase Pl Iness => vesii + ("ss" + a) ; + NCase Pl Elat => vesii + ("st" + a) ; + NCase Pl Illat => vesiin ; + NCase Pl Adess => vesii + ("ll" + a) ; + NCase Pl Ablat => vesii + ("lt" + a) ; + NCase Pl Allat => vesii + "lle" ; + + NPossNom => vete ; + NPossGenPl => Predef.tk 1 vesien ; + NPossTransl Sg => vede + "kse" ; + NPossTransl Pl => vesii + "kse" ; + NPossIllat Sg => Predef.tk 1 veteen ; + NPossIllat Pl => Predef.tk 1 vesiin + } + } ; + +-- A user-friendly variant takes existing forms and infers the vowel harmony. + + mkNoun : (_,_,_,_,_,_,_,_,_,_ : Str) -> CommonNoun = + \talo,talon,talona,taloa,taloon,taloina,taloissa,talojen,taloja,taloihin -> + mkSubst (ifTok Str (Predef.dp 1 talona) "a" "a" "ä") + talo (Predef.tk 1 talon) (Predef.tk 2 talona) taloa taloon + (Predef.tk 2 taloina) (Predef.tk 3 taloissa) talojen taloja taloihin ; + +-- Here some useful special cases; more will be given in $paradigms.Fin.gf$. +-- +-- Nouns with partitive "a"/"ä" ; +-- to account for grade and vowel alternation, three forms are usually enough +-- Examples: "talo", "kukko", "huippu", "koira", "kukka", "syylä",... + + sKukko : (_,_,_ : Str) -> CommonNoun = \kukko,kukon,kukkoja -> + let { + o = Predef.dp 1 kukko ; + a = Predef.dp 1 kukkoja ; + kukkoj = Predef.tk 1 kukkoja ; + i = Predef.dp 1 kukkoj ; + ifi = ifTok Str i "i" ; + kukkoi = ifi kukkoj (Predef.tk 1 kukkoj) ; + e = Predef.dp 1 kukkoi ; + kukoi = Predef.tk 2 kukon + Predef.dp 1 kukkoi + } + in + mkSubst a + kukko + (Predef.tk 1 kukon) + kukko + (kukko + a) + (kukko + o + "n") + (kukkoi + ifi "" "i") + (kukoi + ifi "" "i") + (ifTok Str e "e" (Predef.tk 1 kukkoi + "ien") (kukkoi + ifi "en" "jen")) + kukkoja + (kukkoi + ifi "in" "ihin") ; + +-- The special case with no alternations: the vowel harmony is inferred from the +-- last letter - which must be one of "o", "u", "ö", "y". + + sTalo : Str -> CommonNoun = \talo -> + let {a = getHarmony (Predef.dp 1 talo)} in + sKukko talo (talo + "n") (talo + ("j" + a)) ; + +-- Loan words ending in consonants are actually similar to words like +-- "malli"/"mallin"/"malleja", with the exception that the "i" is not attached +-- to the singular nominative. + + sLinux : Str -> CommonNoun = \linuxia -> + let { + linux = Predef.tk 2 linuxia ; + a = getHarmony (Predef.dp 1 linuxia) ; + linuxi = linux + "i" + } in + mkSubst a + linux + linuxi + linuxi + (linuxi + a) + (linuxi + "in") + (linux + "ei") + (linux + "ei") + (linux + "ien") + (linux + "eja") + (linux + "eihin") ; + +-- Nouns of at least 3 syllables ending with "a" or "ä", like "peruna", "rytinä". + + sPeruna : Str -> CommonNoun = \peruna -> + let { + a = Predef.dp 1 peruna ; + perun = Predef.tk 1 peruna ; + perunoi = perun + (ifTok Str a "a" "o" "ö" + "i") + } + in + mkSubst a + peruna + peruna + peruna + (peruna + a) + (peruna + a + "n") + perunoi + perunoi + (perunoi + "den") + (perunoi + ("t" + a)) + (perunoi + "hin") ; + +-- Surpraisingly, making the test for the partitive, this not only covers +-- "rae", "perhe", "savuke", but also "rengas", "lyhyt" (except $Sg Illat$), etc. + + sRae : (_,_ : Str) -> CommonNoun = \rae,rakeena -> + let { + a = Predef.dp 1 rakeena ; + rakee = Predef.tk 2 rakeena ; + rakei = Predef.tk 1 rakee + "i" ; + raet = rae + (ifTok Str (Predef.dp 1 rae) "e" "t" []) + } + in + mkSubst a + rae + rakee + rakee + (raet + ("t" + a)) + (rakee + "seen") + rakei + rakei + (rakei + "den") + (rakei + ("t" + a)) + (rakei + "siin") ; + + sSusi : (_,_,_ : Str) -> CommonNoun = \susi,suden,sutena -> + let { + a = Predef.dp 1 sutena ; + sude = Predef.tk 1 suden ; + sute = Predef.tk 2 sutena + } + in + mkSubst a + susi + sude + sute + (Predef.tk 1 sute + ("t" + a)) + (sute + "en") + susi + susi + (susi + "en") + (susi + a) + (susi + "in") ; + + sPuu : Str -> CommonNoun = \puu -> + let { + u = Predef.dp 1 puu ; + a = getHarmony u ; + pu = Predef.tk 1 puu ; + pui = pu + "i" + } + in + mkSubst a + puu + puu + puu + (puu + ("t" + a)) + (puu + ("h" + u + "n")) + pui + pui + (pui + "den") + (pui + ("t" + a)) + (pui + "hin") ; + + sSuo : Str -> CommonNoun = \suo -> + let { + o = Predef.dp 1 suo ; + a = getHarmony o ; + soi = Predef.tk 2 suo + (o + "i") + } + in + mkSubst a + suo + suo + suo + (suo + ("t" + a)) + (suo + ("h" + o + "n")) + soi + soi + (soi + "den") + (soi + ("t" + a)) + (soi + "hin") ; + +-- Here in fact it is handy to use the partitive form as the only stem. + + sNainen : Str -> CommonNoun = \naista -> + let { + nainen = Predef.tk 3 naista + "nen" ; + nais = Predef.tk 2 naista ; + naise = nais + "e" ; + naisi = nais + "i" ; + a = Predef.dp 1 naista + } + in + mkSubst a + nainen + naise + naise + (nais + ("t" + a)) + (nais + "een") + naisi + naisi + (nais + "ten") + (nais + ("i" + a)) + (nais + "iin") ; + +-- The following covers: "tilaus", "kaulin", "paimen", "laidun", "sammal", +-- "kyynel" (excep $Sg Iness$ for the last two?). + + sTilaus : (_,_ : Str) -> CommonNoun = \tilaus, tilauksena -> + let { + tilauks = Predef.tk 3 tilauksena ; + tilaukse = tilauks + "e" ; + tilauksi = tilauks + "i" ; + a = Predef.dp 1 tilauksena + } + in + mkSubst a + tilaus + tilaukse + tilaukse + (tilaus + ("t" + a)) + (tilauks + "een") + tilauksi + tilauksi + (tilaus + "ten") + (tilauks + ("i" + a)) + (tilauks + "iin") ; + +-- The following covers nouns like "nauris" and adjectives like "kallis", "tyyris". + + sNauris : (_ : Str) -> CommonNoun = \naurista -> + let { + a = Predef.dp 1 naurista ; + nauris = Predef.tk 2 naurista ; + nauri = Predef.tk 3 naurista ; + naurii = nauri + "i" + } + in + mkSubst a + nauris + naurii + naurii + (nauris + ("t" + a)) + (naurii + "seen") + naurii + naurii + (naurii + "den") + (naurii + ("t" + a)) + (naurii + "siin") ; + +-- The following two are used for adjective comparison. + + sSuurempi : Str -> CommonNoun = \suurempaa -> + let { + a = Predef.dp 1 suurempaa ; + suure = Predef.tk 4 suurempaa ; + suurempi = suure + "mpi" ; + suurempa = suure + ("mp" + a) ; + suuremm = suure + "mm" + } + in + mkSubst a + suurempi + (suuremm + a) + suurempa + (suurempa + a) + (suurempa + (a + "n")) + suurempi + (suuremm + "i") + (suurempi + "en") + (suurempi + a) + (suurempi + "in") ; + + sSuurin : Str -> CommonNoun = \suurinta -> + let { + a = Predef.dp 1 suurinta ; + suuri = Predef.tk 3 suurinta ; + suurin = suuri + "n" ; + suurimma = suuri + ("mm" + a) ; + suurimpa = suuri + ("mp" + a) ; + suurimpi = suuri + "mpi" ; + suurimmi = suuri + "mmi" + } + in + mkSubst a + suurin + suurimma + suurimpa + (suurin + ("t" + a)) + (suurimpa + (a + "n")) + suurimpi + suurimmi + (suurimpi + "en") + (suurimpi + a) + (suurimpi + "in") ; + +-- This auxiliary resolves vowel harmony from a given letter. + +getHarmony : Str -> Str = \u -> + ifTok Str u "a" "a" ( + ifTok Str u "o" "a" ( + ifTok Str u "u" "a" "ä")) ; + + +-- We could use an extension of the following for grade alternation, but we don't; +-- in general, *whether there is* grade alternation must be given in the lexicon +-- anyway (cf. "auto" - "auton", not "audon"). + + weakGrade : Str -> Str = \kukko -> + let { + ku = Predef.tk 3 kukko ; + kk = Predef.tk 1 (Predef.dp 3 kukko) ; + o = Predef.dp 1 kukko ; + ifkk = ifTok Str kk ; + k = + ifkk "kk" "k" ( + ifkk "pp" "p" ( + ifkk "tt" "t" ( + ifkk "nt" "nn" ( + ifkk "mp" "mm" ( + ifkk "rt" "rr" ( + ifkk "lt" "ll" ( + kk))))))) + } + in ku + k + o ; + + +--3 Proper names +-- +-- Proper names are similar to common nouns in the singular. + + mkProperName : CommonNoun -> ProperName = \jussi -> + {s = \\c => jussi.s ! NCase Sg c} ; + +--2 Pronouns +-- +-- Here we define personal and relative pronouns. + + mkPronoun : (_,_,_,_,_ : Str) -> Number -> Person -> Pronoun = + \mina, minun, minua, minuna, minuun, n, p -> + let { + minu = Predef.tk 2 minuna ; + a = Predef.dp 1 minuna + } in + {s = table { + PCase Nom => mina ; + PCase Gen => minun ; + PCase Part => minua ; + PCase Transl => minu + "ksi" ; + PCase Ess => minuna ; + PCase Iness => minu + ("ss" + a) ; + PCase Elat => minu + ("st" + a) ; + PCase Illat => minuun ; + PCase Adess => minu + ("ll" + a) ; + PCase Ablat => minu + ("lt" + a) ; + PCase Allat => minu + "lle" ; + PAcc => Predef.tk 1 minun + "t" + } ; + n = n ; p = p} ; + + pronMina = mkPronoun "minä" "minun" "minua" "minuna" "minuun" Sg P1 ; + pronSina = mkPronoun "sinä" "sinun" "sinua" "sinuna" "sinuun" Sg P2 ; + pronHan = mkPronoun "hän" "hänen" "häntä" "hänenä" "häneen" Sg P3 ; + pronMe = mkPronoun "me" "meidän" "meitä" "meinä" "meihin" Pl P1 ; + pronTe = mkPronoun "te" "teidän" "teitä" "teinä" "teihin" Pl P2 ; + pronHe = mkPronoun "he" "heidän" "heitä" "heinä" "heihin" Pl P3 ; + pronNe = mkPronoun "ne" "niiden" "niitä" "niinä" "niihin" Pl P3 ; + +-- The non-human pronoun "se" ('it') is even more irregular, +-- Its accusative cases do not +-- have a special form with "t", but have the normal genitive/nominative variation. +-- We use the type $ProperName$ for "se", because of the accusative but also +-- because the person and number are as for proper names. + + pronSe : ProperName = { + s = table { + Nom => "se" ; + Gen => "sen" ; + Part => "sitä" ; + Transl => "siksi" ; + Ess => "sinä" ; + Iness => "siinä" ; + Elat => "siitä" ; + Illat => "siihen" ; + Adess => "sillä" ; + Ablat => "siltä" ; + Allat => "sille" + } ; + } ; + +-- The possessive suffixes will be needed in syntax. It will show up +-- as a separate word ("auto &ni"), which needs unlexing. Unlexing also +-- has to fix the vowel harmony in cases like "äiti &nsä". + + suff : Str -> Str = \ni -> "&" + ni ; + + possSuffix : Number => Person => Str = \\n,p => + suff (case of { + => "ni" ; + => "si" ; + => "nsa" ; + => "mme" ; + => "nne" ; + => "nsa" + } ) ; + +-- The relative pronoun, "joka", is inflected in case and number, +-- like common nouns, but it does not take possessive suffixes. +-- The inflextion shows a surprising similarity with "suo". + + relPron : RelPron = + let {jo = sSuo "jo"} in {s = + table { + Sg => table { + Nom => "joka" ; + Gen => "jonka" ; + c => jo.s ! NCase Sg c + } ; + Pl => table { + Nom => "jotka" ; + c => "j" + (jo.s ! NCase Pl c) + } + } + } ; + + mikaInt : Number => Case => Str = + let { + mi = sSuo "mi" + } in + table { + Sg => table { + Nom => "mikä" ; + Gen => "minkä" ; + c => mi.s ! NCase Sg c + } ; + Pl => table { + Nom => "mitkä" ; + Gen => "mittenkä" ; + c => mi.s ! NCase Sg c + } + } ; + + kukaInt : Number => Case => Str = + let { + ku = sRae "kuka" "kenenä" ; + ket = sRae "kuka" "keinä"} in + table { + Sg => table { + Nom => "kuka" ; + Part => "ketä" ; + Illat => "keneen" ; + c => ku.s ! NCase Sg c + } ; + Pl => table { + Nom => "ketkä" ; + Illat => "keihin" ; + c => ket.s ! NCase Pl c + } + } ; + +caseTable : Number -> CommonNoun -> Case => Str = \n,cn -> + \\c => cn.s ! NCase n c ; + + +--2 Adjectives +-- +-- For the comparison of adjectives, three noun declensions +-- are needed in the worst case. + + mkAdjDegr : (_,_,_ : CommonNoun) -> AdjDegr = \hyva,parempi,paras -> + {s = table { + Pos => hyva.s ; + Comp => parempi.s ; + Sup => paras.s + } + } ; + +-- However, it is usually enough to give the positive declension and +-- the characteristic forms of comparative and superlative. + + regAdjDegr : CommonNoun -> Str -> Str -> AdjDegr = \kiva, kivempaa, kivinta -> + mkAdjDegr kiva (sSuurempi kivempaa) (sSuurin kivinta) ; + + +--3 Verbs +-- + + mkVerb : (_,_,_,_,_ : Str) -> Verb = \tulla,tulen,tulee,tulevat,tulkaa -> + let { + tule = Predef.tk 1 tulen ; + a = Predef.dp 1 tulkaa + } in + {s = table { + Inf => tulla ; + Ind Sg P1 => tulen ; + Ind Sg P2 => tule + "t" ; + Ind Sg P3 => tulee ; + Ind Pl P1 => tule + "mme" ; + Ind Pl P2 => tule + "tte" ; + Ind Pl P3 => tulevat ; + Imper Sg => tule ; + Imper Pl => tulkaa ; + ImpNegPl => Predef.tk 2 tulkaa + (ifTok Str a "a" "o" "ö") + } + } ; + +-- For "sanoa", "valua", "kysyä". + + vSanoa : Str -> Verb = \sanoa -> + let { + a = Predef.dp 1 sanoa ; + sano = Predef.tk 1 sanoa ; + o = Predef.dp 1 sano + } in + mkVerb + sanoa + (sano + "n") + (sano + o) + (sano + (("v" + a) + "t")) + (sano + (("k" + a) + a)) ; + +-- For "ottaa", "käyttää", "löytää", "huoltaa", "hiihtää", "siirtää". + + vOttaa : (_,_ : Str) -> Verb = \ottaa,otan -> + let { + a = Predef.dp 1 ottaa ; + ota = Predef.tk 1 otan ; + otta = Predef.tk 1 ottaa + } in + mkVerb + ottaa + (ota + "n") + ottaa + (otta + (("v" + a) + "t")) + (otta + (("k" + a) + a)) ; + +-- For "poistaa", "ryystää". + + vPoistaa : Str -> Verb = \poistaa -> + vOttaa poistaa (Predef.tk 1 poistaa + "n") ; + +-- For "juosta", "piestä", "nousta", "rangaista", "kävellä", "surra", "panna". + + vJuosta : (_,_ : Str) -> Verb = \juosta,juoksen -> + let { + a = Predef.dp 1 juosta ; + juokse = Predef.tk 1 juoksen ; + juos = Predef.tk 2 juosta + } in + mkVerb + juosta + juoksen + (juokse + "e") + (juokse + (("v" + a) + "t")) + (juos + (("k" + a) + a)) ; + +-- For "juoda", "syödä". + + vJuoda : Str -> Verb = \juoda -> + let { + a = Predef.dp 1 juoda ; + juo = Predef.tk 2 juoda + } in + mkVerb + juoda + (juo + "n") + juo + (juo + (("v" + a) + "t")) + (juo + (("k" + a) + a)) ; + + + verbOlla : Verb = mkVerb "olla" "olen" "on" "ovat" "olkaa" ; + +-- The negating operator "ei" is actually a verb, which has has present +-- indicative and imperative forms, but no infinitive. + + verbEi : Verb = + let {ei = mkVerb nonExist "en" "ei" "eivät" "älkää"} in + {s = table { + Ind Pl P3 => "eivät" ; + v => ei.s ! v + } + } ; + + +--2 Some structural words + + kuinConj = "kuin" ; + + conjEtta = "että" ; + advSiten = "siten" ; + + mikakukaInt : Gender => Number => Case => Str = + table { + NonHuman => mikaInt ; + Human => kukaInt + } ; + + kaikkiPron : Case => Str = + let {kaiket = caseTable Pl (sKukko "kaikki" "kaiken" "kaikkia")} in + table { + Nom => "kaikki" ; + c => kaiket ! c + } ; + + stopPunct = "." ; + commaPunct = "," ; + questPunct = "?" ; + exclPunct = "!" ; + + koPart = suff "ko" ; + +} ; \ No newline at end of file diff --git a/grammars/resource/finnish/ResFin.gf b/grammars/resource/finnish/ResFin.gf new file mode 100644 index 000000000..5493f7ac7 --- /dev/null +++ b/grammars/resource/finnish/ResFin.gf @@ -0,0 +1,215 @@ +--1 The Top-Level Finnish Resource Grammar +-- +-- Aarne Ranta 2002 -- 2003 +-- +-- This is the Finnish concrete syntax of the multilingual resource +-- grammar. Most of the work is done in the file $syntax.Fin.gf$. +-- However, for the purpose of documentation, we make here explicit the +-- linearization types of each category, so that their structures and +-- dependencies can be seen. +-- Another substantial part are the linearization rules of some +-- structural words. +-- +-- The users of the resource grammar should not look at this file for the +-- linearization rules, which are in fact hidden in the document version. +-- They should use $resource.Abs.gf$ to access the syntactic rules. +-- This file can be consulted in those, hopefully rare, occasions in which +-- one has to know how the syntactic categories are +-- implemented. The parameter types are defined in $types.Fin.gf$. + +concrete ResFin of ResAbs = open Prelude, Syntax in { + +flags + startcat=Phr ; + parser=chart ; + +lincat + N = CommNoun ; + -- = {s : NForm => Str ; g : Gender} + CN = CommNounPhrase ; + NP = {s : NPForm => Str ; n : Number ; p : NPPerson} ; + PN = {s : Case => Str} ; + Det = {s : Gender => Case => Str ; n : Number} ; + Fun = Function ; + -- = CommNounPhrase ** {c : NPForm} ; + Fun2 = Function ** {c2 : NPForm} ; + + Adj1 = Adjective ; + -- = CommonNoun + Adj2 = Adjective ** {c : NPForm} ; + AdjDeg = {s : Degree => NForm => Str} ; + AP = {s : AdjPos => Number => Case => Str} ; + + V = Verb ; + -- = {s : VForm => Str} + VP = Verb ** {s2 : VForm => Str} ; + TV = TransVerb ; + -- = Verb ** {s3, s4 : Str ; c : ComplCase} ; + V3 = TransVerb ** {s5, s6 : Str ; c2 : ComplCase} ; + VS = Verb ; + + AdV = {s : Str} ; + + S = Sentence ; + -- = {s : Str} ; + Slash = Sentence ** {s2 : Str ; c : Case} ; + + RP = {s : Number => Case => Str} ; + RC = {s : Number => Str} ; + + IP = {s : NPForm => Str ; n : Number} ; + Qu = {s : Str} ; + Imp = {s : Number => Str} ; + Phr = {s : Str} ; + + Conj = {s : Str ; n : Number} ; + ConjD = {s1 : Str ; s2 : Str ; n : Number} ; + + ListS = {s1 : Str ; s2 : Str} ; + ListAP = {s1,s2 : AdjPos => Number => Case => Str} ; + ListNP = {s1,s2 : NPForm => Str ; n : Number ; p : NPPerson} ; + +--. + +lin + UseN = noun2CommNounPhrase ; + ModAdj = modCommNounPhrase ; + ModGenOne = npGenDet singular ; + ModGenMany = npGenDet plural ; + UsePN = nameNounPhrase ; + UseFun = funAsCommNounPhrase ; + AppFun = appFunComm ; + AppFun2 = appFun2 ; + AdjP1 = adj2adjPhrase ; + ComplAdj = complAdj ; + PositAdjP = positAdjPhrase ; + ComparAdjP = comparAdjPhrase ; + SuperlNP = superlNounPhrase ; + + DetNP = detNounPhrase ; + IndefOneNP = indefNounPhrase singular ; + IndefManyNP = indefNounPhrase plural ; + DefOneNP = defNounPhrase singular ; + DefManyNP = defNounPhrase plural ; + + PredVP = predVerbPhrase ; + PosV = predVerb True ; + NegV = predVerb False ; + PosA = predAdjective True ; + NegA = predAdjective False ; + PosCN = predCommNoun True ; + NegCN = predCommNoun False ; + PosTV = complTransVerb True ; + NegTV = complTransVerb False ; + PosV3 = complDitransVerb True ; + NegV3 = complDitransVerb False ; + PosNP = predNounPhrase True ; + NegNP = predNounPhrase False ; + PosVS = complSentVerb True ; + NegVS = complSentVerb False ; + VTrans = transAsVerb ; + + AdvVP = adVerbPhrase ; + LocNP = locativeNounPhrase ; + AdvCN = advCommNounPhrase ; + AdvAP = advAdjPhrase ; + + PosSlashTV = slashTransVerb True ; + NegSlashTV = slashTransVerb False ; + + IdRP = identRelPron ; + FunRP = funRelPron ; + RelVP = relVerbPhrase ; + RelSlash = relSlash ; + ModRC = modRelClause ; + RelSuch = relSuch ; + + WhoOne = intPronWho singular ; + WhoMany = intPronWho plural ; + WhatOne = intPronWhat singular ; + WhatMany = intPronWhat plural ; + FunIP = funIntPron ; + NounIPOne = nounIntPron singular ; + NounIPMany = nounIntPron plural ; + + QuestVP = questVerbPhrase ; + IntVP = intVerbPhrase ; + IntSlash = intSlash ; + QuestAdv = questAdverbial ; + + ImperVP = imperVerbPhrase ; + + IndicPhrase = indicUtt ; + QuestPhrase = interrogUtt ; + ImperOne = imperUtterance singular ; + ImperMany = imperUtterance plural ; + + AdvS = advSentence ; + +lin + TwoS = twoSentence ; + ConsS = consSentence ; + ConjS = conjunctSentence ; + ConjDS = conjunctDistrSentence ; + + TwoAP = twoAdjPhrase ; + ConsAP = consAdjPhrase ; + ConjAP = conjunctAdjPhrase ; + ConjDAP = conjunctDistrAdjPhrase ; + + TwoNP = twoNounPhrase ; + ConsNP = consNounPhrase ; + ConjNP = conjunctNounPhrase ; + ConjDNP = conjunctDistrNounPhrase ; + + SubjS = subjunctSentence ; + SubjImper = subjunctImperative ; + SubjQu = subjunctQuestion ; + + PhrNP = useNounPhrase ; + PhrOneCN = useCommonNounPhrase singular ; + PhrManyCN = useCommonNounPhrase plural ; + PhrIP ip = ip ; + PhrIAdv ia = ia ; + + OnePhr p = p ; + ConsPhr = cc2 ; + +lin + INP = pronNounPhrase pronMina ; + ThouNP = pronNounPhrase pronSina ; + HeNP = pronNounPhrase pronHan ; + SheNP = pronNounPhrase pronHan ; + ItNP = nameNounPhrase pronSe ; + WeNP = pronNounPhrase pronMe ; + YeNP = pronNounPhrase pronTe ; + YouNP = pronNounPhrase pronTe ; + TheyNP = pronNounPhrase pronHe ; --- ne + + EveryDet = jokainenDet ; + AllDet = kaikkiDet ; + WhichDet = mikaDet ; + MostDet = useimmatDet ; + + HowIAdv = ss "kuinka" ; + WhenIAdv = ss "koska" ; + WhereIAdv = ss "missä" ; + WhyIAdv = ss "miksi" ; + + AndConj = ss "ja" ** {n = Pl} ; + OrConj = ss "tai" ** {n = Sg} ; + BothAnd = sd2 "sekä" "että" ** {n = Pl} ; + EitherOr = sd2 "joko" "tai" ** {n = Sg} ; + NeitherNor = sd2 "ei" "eikä" ** {n = Sg} ; + IfSubj = ss "jos" ; + WhenSubj = ss "kun" ; + + PhrYes = ss ("Kyllä" ++ stopPunct) ; + PhrNo = ss ("Ei" ++ stopPunct) ; + + VeryAdv = ss "hyvin" ; + TooAdv = ss "liian" ; + OtherwiseAdv = ss "muuten" ; + ThereforeAdv = ss "siksi" ; + +} ; \ No newline at end of file diff --git a/grammars/resource/finnish/Syntax.gf b/grammars/resource/finnish/Syntax.gf new file mode 100644 index 000000000..66977b610 --- /dev/null +++ b/grammars/resource/finnish/Syntax.gf @@ -0,0 +1,974 @@ +--1 A Small Finnish Resource Syntax +-- +-- Aarne Ranta 2003 +-- +-- This resource grammar contains definitions needed to construct +-- indicative, interrogative, and imperative sentences in Finnish. +-- +-- The following files are presupposed: + +resource Syntax = Morpho ** open Prelude, (CO = Coordination) in { + +--2 Common Nouns +-- +-- Simple common nouns are defined as the type $CommNoun$ in $morpho.Fin.gf$. + +--3 Common noun phrases + +-- In Finnish, common noun phrases behave like simple common nouns, except that +-- we need a kind of a *gender* parameter telling if the noun is human or not. +-- This parameter regulates determiners such as "joku"/"jokin" ('some') and +-- "kuka"/"mikä" ('which'). +-- +-- A subtle reason forces us to distinguish the parameters of common noun phrases +-- from those of morphological common nouns: the parameter value $NPossNom$ is +-- syntactically applicable to each of $Sg Nom$, $Pl Nom$, $Sg Gen$. In morphology, +-- these forms are always the same ("autoni"), but with complex common nouns, we +-- have three different forms: "iso autoni", "isot autoni", "ison autoni". + +oper + CommNoun = {s : NForm => Str ; g : Gender} ; + + CommNounPhrase = {s : Bool => Number => Case => Str ; g : Gender} ; + + noun2CommNounPhrase : CommNoun -> CommNounPhrase = \man -> + useCN man ** {g = man.g} ; + + n2n = noun2CommNounPhrase ; + + useCN : CommonNoun -> {s : Bool => Number => Case => Str} = \auto -> + {s = table { + True => \\n,c => case of { + <_, Nom> => auto.s ! NPossNom ; + => auto.s ! NPossNom ; + => auto.s ! NPossGenPl ; + <_,Transl> => auto.s ! NPossTransl n ; + <_,Illat> => auto.s ! NPossIllat n ; + _ => auto.s ! NCase n c + } ; + False => \\n,c => auto.s ! NCase n c + } + } ; + + cnNoHum : CommonNoun -> CommNoun = \cn -> cn ** {g = NonHuman} ; + cnHum : CommonNoun -> CommNoun = \cn -> cn ** {g = Human} ; + +--2 Noun phrases +-- +-- Two forms of *virtual accusative* are needed for nouns in singular, +-- the nominative and the genitive one ("ostan talon"/"osta talo"). +-- For nouns in plural, only a nominative accusative exist. Pronouns +-- have a uniform, special accusative form ("minut", etc). + +param + NPForm = NPCase Case | NPAccNom | NPAccGen ; + +-- The *person* of a noun phrase is also special, to steer the use of +-- possessive suffixes. It expresses a distinction between pronominal and +-- non-pronominal noun phrases. The pronominal ones impose possessive suffixes +-- in genitival constructions ("minun taloni", "hänen talonsa"), the non-pronominal +-- ones don't ("Jussin talo"). As for verbal agreement, non-pronominal noun +-- phrases are third-person. + + NPPerson = NP3 | NPP Person ; + +oper + np2Person : NPPerson -> Person = \n -> case n of { + NP3 => P3 ; + NPP p => p + } ; + +oper + npForm2Case : Number -> NPForm -> Case = \n,f -> case f of { + NPCase c => c ; + NPAccNom => Nom ; + NPAccGen => case n of { + Sg => Gen ; + Pl => Nom + } + } ; + + npForm2PForm : NPForm -> PForm = \f -> case f of { + NPCase c => PCase c ; + _ => PAcc + } ; + + NounPhrase : Type = {s : NPForm => Str ; n : Number ; p : NPPerson} ; + + nameNounPhrase : ProperName -> NounPhrase = \jussi -> + {s = \\f => jussi.s ! npForm2Case Sg f ; n = Sg ; p = NP3} ; + + singularNounPhrase : CommNounPhrase -> NounPhrase = \cn -> + {s = \\f => cn.s ! False ! Sg ! (npForm2Case Sg f) ; n = Sg ; p = NP3} ; + + pluralNounPhrase : CommNounPhrase -> NounPhrase = \cn -> + {s = \\f => cn.s ! False ! Pl ! (npForm2Case Pl f) ; n = Pl ; p = NP3} ; + + pronNounPhrase : Pronoun -> NounPhrase = \pron -> + {s = \\f => pron.s ! npForm2PForm f ; n = pron.n ; p = NPP pron.p} ; + +-- *Partitive noun phrases* use the partitive instead of the nominative +-- and accusative forms. + + npForm2CasePart : NPForm -> Case = \f -> case f of { + NPCase Nom => Part ; + NPCase c => c ; + _ => Part + } ; + + partNounPhrase : Number -> CommNounPhrase -> NounPhrase = \n, cn -> + {s = \\f => cn.s ! False ! n ! (npForm2CasePart f) ; n = n ; p = NP3} ; + + + +--2 Determiners +-- +-- Most determiners are inflected like nouns. They have an inherent number +-- that is given to the noun that is being determined. + + Determiner : Type = {s : Gender => Case => Str ; n : Number} ; + + detNounPhrase : Determiner -> CommNounPhrase -> NounPhrase = \joku, mies -> + {s = \\f => let {c = npForm2Case joku.n f} in + joku.s ! mies.g ! c ++ mies.s ! False ! joku.n ! c ; + n = joku.n ; + p = NP3 + } ; + + mkDeterminerGen : Number -> (_,_ : Case => Str) -> Determiner = \n,mika,kuka -> + {s = table { + NonHuman => mika ; + Human => kuka + } ; + n = n + } ; + + mkDeterminer : Number -> (Case => Str) -> Determiner = \n,kaikki -> + mkDeterminerGen n kaikki kaikki ; + + jokainenDet = mkDeterminer Sg (caseTable Sg (sNainen "jokaista")) ; + kaikkiDet = mkDeterminer Pl kaikkiPron ; + useimmatDet = mkDeterminer Pl (caseTable Pl (sSuurin "useinta")) ; + mikaDet = mkDeterminerGen Sg (mikaInt ! Sg) (kukaInt ! Sg) ; + mitkaDet = mkDeterminerGen Pl (mikaInt ! Pl) (kukaInt ! Pl) ; + + indefNounPhrase : Number -> CommNounPhrase -> NounPhrase = \n,mies -> + case n of { + Sg => singularNounPhrase mies ; + Pl => partNounPhrase plural mies + } ; + + defNounPhrase : Number -> CommNounPhrase -> NounPhrase = \n,mies -> + case n of { + Sg => singularNounPhrase mies ; + Pl => pluralNounPhrase mies + } ; + + +-- Genitives of noun phrases can be used like determiners, to build noun phrases. +-- The number argument makes the difference between "Jussin talo" - "Jussin talot". +-- The NP person of the 'owner' decides if there is a possessive suffix. + + npGenDet : Number -> NounPhrase -> CommNounPhrase -> NounPhrase = \n,jussi,talo -> + {s = \\c => jussi.s ! NPCase Gen ++ + ifPossSuffix talo jussi.p jussi.n (npForm2Case n c) ; + n = n ; + p = NP3 + } ; + + ifPossSuffix : CommNounPhrase -> NPPerson -> Number -> Case -> Str = + \talo,np,n,c -> case np of { + NP3 => talo.s ! False ! n ! c ; + NPP p => talo.s ! True ! n ! c ++ possSuffix ! n ! p + } ; + +-- *Bare plural noun phrases*, like "koivut" in "koivut ovat valkoisia", +-- are similar to definite plurals. + + plurDet : CommNounPhrase -> NounPhrase = pluralNounPhrase ; + + +--2 Adjectives +-- +-- Adjectival phrases are used either as attributes or in predicative position. +-- In the attributive position, all cases occur; in the predicative position, only +-- the nominative, partitive, translative, and essive - but we ignore this +-- restriction for simplicity. The important thing with the parameter is to +-- regulate the word order of complex adjectival phrases: cf. predicative +-- "(kuusi on) jaollinen kolmella" vs. attributive "kolmella jaollinen (luku)". +-- In comparatives, the whole construction is affected: "suurempi kuin kolme" +-- vs. "kolmea suurempi". (Actually, in the predicative position, the two +-- are in free variation, the distinguished one being the normal choice: +-- "kuusi on kolmella jaollinen" is possible, but not quite neutral.) + +param + AdjPos = APred | AAttr ; + +oper + AdjPhrase : Type = {s : AdjPos => Number => Case => Str} ; + + adj2adjPhrase : Adjective -> AdjPhrase = \uusi -> + {s = \\_,n,c => uusi.s ! NCase n c} ; + + +--3 Comparison adjectives +-- +-- Each of the comparison forms has a characteristic use: +-- +-- Positive forms are used alone, as adjectival phrases ("iso"). + + positAdjPhrase : AdjDegr -> AdjPhrase = \iso -> + adj2adjPhrase {s = iso.s ! Pos} ; + +-- Comparative forms are used with an object of comparison, as +-- adjectival phrases ("isompi kuin te"/"teitä isompi"). + + comparAdjPhrase : AdjDegr -> NounPhrase -> AdjPhrase = \iso, te -> + {s = let {teitaisompi : Number => Case => Str = + \\n,c => te.s ! NPCase Part ++ iso.s ! Comp ! NCase n c} in + table { + APred => variants { + \\n,c => iso.s ! Comp ! NCase n c ++ kuinConj ++ te.s ! NPCase Nom ; + teitaisompi + } ; + AAttr => teitaisompi + } + } ; + +-- Superlative forms are used with a modified noun, picking out the +-- maximal representative of a domain ("isoin talo"). + + superlNounPhrase : AdjDegr -> CommNounPhrase -> NounPhrase = \iso,talo -> + {s = \\np => let {c = npForm2Case Sg np} in + iso.s ! Sup ! NCase Sg c ++ talo.s ! False ! Sg ! c ; + n = Sg ; + p = NP3 + } ; + +--3 Two-place adjectives +-- +-- A two-place adjective is an adjective with a case used after (or before) +-- the complement. The case can be the genitival accusative, which is different +-- in the singular and the plural ("rajan ylittävä"/"rajat ylittävä"). +-- The order of the adjective and its argument depends on the case: the local +-- cases favour Adj + Noun in the predicative position ("hyvä painissa", +-- "tyytyväinen vaalitulokseen", "jaollinen kolmella"), which is not a possible +-- order for the accusative case. + + AdjCompl = Adjective ** {c : NPForm} ; + + complAdj : AdjCompl -> NounPhrase -> AdjPhrase = \hyva,paini -> + let { + hyvat : Number => Case => Str = \\n,c => hyva.s ! NCase n c ; + painissa : Str = paini.s ! hyva.c + } + in + {s = table { + AAttr => \\n,c => painissa ++ hyvat ! n ! c ; + APred => \\n,c => if_then_else Str + (isLocalNPForm hyva.c) + (variants { + hyvat ! n ! c ++ painissa ; + painissa ++ hyvat ! n ! c + } + ) + (painissa ++ hyvat ! n ! c) + } + } ; + + isLocalNPForm : NPForm -> Bool = \c -> case c of { + NPCase Iness => True ; + NPCase Elat => True ; + NPCase Illat => True ; + NPCase Adess => True ; + NPCase Ablat => True ; + NPCase Allat => True ; + _ => False + } ; + + +--3 Modification of common nouns +-- +-- The two main functions of adjective are in predication ("Jussi on iso") +-- and in modification ("iso mies"). Predication will be defined +-- later, in the chapter on verbs. +-- +-- Modification uses the attributive form of an adjectival phrase. +-- The adjective always comes before the noun. The possessive suffix is +-- given to the noun. + + modCommNounPhrase : AdjPhrase -> CommNounPhrase -> CommNounPhrase = \iso,mies -> + {s = \\p,n,c => iso.s ! AAttr ! n ! c ++ mies.s ! p ! n ! c ; + g = mies.g + } ; + +--2 Function expressions + +-- A function expression is a common noun together with the +-- case taken by its argument ("x'n vaimo"). +-- The type is analogous to two-place adjectives and transitive verbs; +-- but here the genitive is by far the commonest case. The possessive suffix +-- is then needed with pronominal arguments. + + Function = CommNounPhrase ** {c : NPForm} ; + +-- The application of a function gives, in the first place, a common noun: +-- "Jussi vaimo/vaimot". From this, other rules of the resource grammar +-- give noun phrases, such as "Jussi vaimo", "Jussin vaimot", +-- "Jussin ja Marin äidit", and "Jussin ja Marin äiti" (the +-- latter two corresponding to distributive and collective functions, +-- respectively). Semantics will eventually tell when each +-- of the readings is meaningful. + + appFunComm : Function -> NounPhrase -> CommNounPhrase = \vaimo, jussi -> + {s = \\p,n,c => case vaimo.c of { + NPCase Gen => jussi.s ! NPCase Gen ++ + ifPossSuffix vaimo jussi.p jussi.n c ; + h => vaimo.s ! False ! n ! c ++ jussi.s ! h + } ; + g = vaimo.g + } ; + +-- Notice the switched word order in other cases than the genitive, e.g. +-- "veli Jussille". +-- +-- It is possible to use a function word as a common noun; the semantics is +-- often existential or indexical. + + funAsCommNounPhrase : Function -> CommNounPhrase = \x -> x ; + +-- The following is an aggregate corresponding to function application +-- producing "John's mother" and "the mother of John". It does not appear in the +-- resource grammar API as a primitive. + + appFun : Bool -> Function -> NounPhrase -> NounPhrase = \coll, vaimo,jussi -> + let {n = jussi.n ; nf = if_then_else Number coll Sg n} in + npGenDet nf jussi vaimo ; + +-- The commonest case is functions with the genitive case. + + funGen : CommNounPhrase -> Function = \vaimo -> + vaimo ** {c = NPCase Gen} ; + +-- Two-place functions add one argument place. + + Function2 = Function ** {c2 : NPForm} ; + +-- There application starts by filling the first place. + + appFun2 : Function2 -> NounPhrase -> Function = \juna, turku -> + {s = \\p,n,c => juna.s ! False ! n ! c ++ turku.s ! juna.c ; + g = juna.g ; + c = juna.c2 + } ; + + +--2 Verbs +-- +--3 Verb phrases +-- +-- Verb phrases are discontinuous: the two parts of a verb phrase are +-- (s) an inflected verb, (s2) a complement. +-- For instance: "on" - "kaunis" ; "ei" - "ole kaunis" ; "sisältää" - "rikkiä". + + VerbPhrase = Verb ** {s2 : VForm => Str} ; + +-- From the inflection table, we select the finite form as function +-- of person and number: + + indicVerb : Verb -> Person -> Number -> Str = \v,p,n -> + v.s ! Ind n p ; + +-- A simple verb can be made into a verb phrase with an empty complement, e.g. +-- "ui" - []. +-- There are two versions, depending on if we want to negate the verb. +-- In the negated form, the negative verb "ei" becomes the verb, and the +-- complement is a special infinite form of the verb (usually similar to the +-- 2nd person singular imperative): "ei" - "ui". +-- +-- N.B. negation is *not* a function applicable to a verb phrase, since +-- double negations with "ei" are not grammatical. + + predVerb : Bool -> Verb -> VerbPhrase = \b,walk -> + let { + noCompl : {s2 : VForm => Str} = {s2 = \\_ => []} ; + infCompl : {s2 : VForm => Str} = {s2 = table { + Imper Pl => walk.s ! ImpNegPl ; + _ => walk.s ! vFormNeg + } + } + } + in if_then_else VerbPhrase b (walk ** noCompl) (verbEi ** infCompl) ; + +-- (N.B. local definitions workaround for poor type inference in GF 1.2). + +-- Sometimes we want to extract the verb part of a verb phrase. Not strictly +-- necessary since this is a consequence of record subtyping. + + verbOfPhrase : VerbPhrase -> Verb = \v -> {s = v.s} ; + +-- Verb phrases can also be formed from adjectives ("on vanha"), +-- common nouns ("on mies"), and noun phrases ("on Jussi"). +-- The third rule is overgenerating: "on jokainen mies" has to be ruled out +-- on semantic grounds. +-- +-- For adjectives and common nouns, notice the case difference in the complement +-- depending on number: "on kaunis" - "ovat kauniita". We ignore the forms +-- "on kaunista", used with mass terms, and "ovat kauniit", used in +-- constructions of the "plurale tantum" kind. The adjective rule can be defined +-- in terms of the common noun rule. + + predAdjective : Bool -> AdjPhrase -> VerbPhrase = \b,iso -> + let {isot : CommNounPhrase = {s = \\_ => iso.s ! APred ; g = NonHuman}} + in predCommNoun b isot ; + + predCommNoun : Bool -> CommNounPhrase -> VerbPhrase = \b,mies -> + let { + miehia : VForm => Str = \\v => case vform2number v of { + Sg => mies.s ! False ! Sg ! Nom ; + Pl => mies.s ! False ! Pl ! Part + } ; + olemiehia : VForm => Str = + \\v => verbOlla.s ! vFormNeg ++ miehia ! v + } + in if_then_else VerbPhrase b + (verbOlla ** {s2 = miehia}) + (verbEi ** {s2 = olemiehia}) ; + + predNounPhrase : Bool -> NounPhrase -> VerbPhrase = \b,jussi -> + let {jussia : Bool => Number => Case => Str = \\_,_,_ => jussi.s ! NPCase Nom} + in predCommNoun b {s = jussia ; g = Human} ; --- gender does not matter + + +--3 Transitive verbs +-- +-- Transitive verbs are verbs with a case and, possibly, a preposition +-- or a postposition for the complement, +-- in analogy with two-place adjectives and functions. +-- One might prefer to use the term "2-place verb", since +-- "transitive" traditionally means that the inherent preposition is empty. +-- Such a verb is one with a *direct object*. + +param + ComplCase = CCase Case | CAcc ; + +oper + TransVerb : Type = Verb ** {s3, s4 : Str ; c : ComplCase} ; + +-- The rule for using transitive verbs is the complementization rule. +-- +-- N.B. One or both of the pre- and postposition are empty. + + complTransVerb : Bool -> TransVerb -> NounPhrase -> VerbPhrase = \b,ostaa,talo -> + let { + ostan = predVerb b ostaa ; + talon : VForm => Str = \\v => + ostaa.s3 ++ talo.s ! complementCase b ostaa.c v ++ ostaa.s4 + } + in { + s = ostan.s ; + s2 = \\v => ostan.s2 ! v ++ talon ! v + } ; + +-- N.B. If the case is accusative, it becomes partitive in negated verb phrases. +-- The choice between the nominative and genitive accusatives depends on the verb +-- form. + + complementCase : Bool -> ComplCase -> VForm -> NPForm = \b,c,v -> case c of { + CCase k => NPCase k ; + CAcc => case b of { + True => case v of { + Inf => NPAccNom ; + Ind _ _ => NPAccGen ; + Imper _ => NPAccNom ; + ImpNegPl => NPCase Part + } ; + _ => NPCase Part + } + } ; + +-- Verbs that take their object with a case other than the accusative, +-- without pre- or postposition: + + mkTransVerbCase : Verb -> Case -> TransVerb = \nauraa,c -> + nauraa ** {s3 = [] ; s4 = [] ; c = CCase c} ; + +-- Verbs that take direct object with the accusative: + + mkTransVerbDir : Verb -> TransVerb = \ostaa -> + ostaa ** {s3 = [] ; s4 = [] ; c = CAcc} ; + +-- Transitive verbs can be used elliptically as verbs. The semantics +-- is left to applications. The definition is trivial, due to record +-- subtyping. + + transAsVerb : TransVerb -> Verb = \juoda -> + juoda ; + +-- *Ditransitive verbs* are verbs with three argument places. +-- We treat so far only the rule in which the ditransitive +-- verb takes both complements to form a verb phrase. + + DitransVerb = TransVerb ** {s5, s6 : Str ; c2 : ComplCase} ; + + complDitransVerb : + Bool -> DitransVerb -> NounPhrase -> NounPhrase -> VerbPhrase = + \b,ostaa,talo,me -> + let { + ostan = predVerb b ostaa ; + talon : VForm => Str = \\v => + ostaa.s3 ++ talo.s ! complementCase b ostaa.c v ++ ostaa.s4 ; + meille : VForm => Str = \\v => + ostaa.s5 ++ me.s ! complementCase b ostaa.c2 v ++ ostaa.s6 + } + in { + s = ostan.s ; + s2 = \\v => ostan.s2 ! v ++ talon ! v ++ meille ! v + } ; + + +--2 Adverbials +-- +-- Adverbials are not inflected (we ignore comparison, and treat +-- compared adverbials as separate expressions; this could be done another way). + + Adverb : Type = SS ; + +-- This rule adds the adverbial as a prefix or a suffix to the complement, +-- in free variation. + + adVerbPhrase : VerbPhrase -> Adverb -> VerbPhrase = \laulaa, hyvin -> + {s = laulaa.s ; + s2 = \\v => bothWays (laulaa.s2 ! v) hyvin.s + } ; + + advAdjPhrase : Adverb -> AdjPhrase -> AdjPhrase = \liian, iso -> + {s = \\p,n,c => liian.s ++ iso.s ! p ! n ! c + } ; + +-- Adverbials are typically generated by case, prepositions, or postpositions. +-- The rule for creating locative noun phrases by the inessive case +-- is a shaky, since the adessive is often required. + + prepPhrase : Str -> Case -> NounPhrase -> Adverb = \ennen,c,talvi -> + ss (ennen ++ talvi.s ! NPCase c) ; + + postpPhrase : Str -> Case -> NounPhrase -> Adverb = \aikana,c,talvi -> + ss (talvi.s ! NPCase c ++ aikana) ; + + caseAdv : Case -> NounPhrase -> Adverb = prepPhrase [] ; + + locativeNounPhrase : NounPhrase -> Adverb = \np -> --- caseAdv Iness ; + ss (np.s ! NPCase Iness) ; + +-- This is a source of the "mann with a telescope" ambiguity, and may produce +-- strange things, like "autot aina" (while "autot tänään" is OK). +-- Semantics will have to make finer distinctions among adverbials. + + advCommNounPhrase : CommNounPhrase -> Adverb -> CommNounPhrase = \auto,nyt -> + {s = \\b,n,c => auto.s ! b ! n ! c ++ nyt.s ; + g = auto.g + } ; + +--2 Sentences +-- +-- Sentences are not inflected in this fragment of Finnish without tense. + + Sentence : Type = SS ; + +-- This is the traditional $S -> NP VP$ rule. It takes care of +-- agreement between subject and verb. Recall that the VP may already +-- contain negation. + + predVerbPhrase : NounPhrase -> VerbPhrase -> Sentence = \jussi,uida -> + let {p = np2Person jussi.p} in + ss (jussi.s ! NPCase Nom ++ uida.s ! Ind jussi.n p ++ uida.s2 ! Ind jussi.n p) ; + +-- This is a macro for simultaneous predication and complementization. + + predTransVerb : Bool -> NounPhrase -> TransVerb -> NounPhrase -> Sentence = + \b,you,see,john -> + predVerbPhrase you (complTransVerb b see john) ; + +--3 Sentence-complement verbs +-- +-- Sentence-complement verbs take sentences as complements. + + SentenceVerb : Type = Verb ; + +-- To generate "sanoo että Jussi ui" / "ei sano että Jussi ui" + + complSentVerb : Bool -> SentenceVerb -> Sentence -> VerbPhrase = + \b,sanoa,jussiui -> + let { + sanon = predVerb b sanoa + } + in { + s = sanon.s ; + s2 = \\v => sanon.s2 ! v ++ conjEtta ++ jussiui.s + } ; + + +--2 Sentences missing noun phrases +-- +-- This is one instance of Gazdar's *slash categories*, corresponding to his +-- $S/NP$. +-- We cannot have - nor would we want to have - a productive slash-category former. +-- Perhaps a handful more will be needed. +-- +-- Notice that the slash category has a similar relation to sentences as +-- transitive verbs have to verbs: it's like a *sentence taking a complement*. +-- +-- Interestingly, the distinction between prepositions and postpositions +-- neutralizes: even prepositions are attached after relative and interrogative +-- pronouns: "jota ennen" cf. "ennen talvea". Otherwise, the category and +-- the rules are very similar to transitive verbs. Notice that the case gets +-- fixed by the Boolean parameter and the subject. + + SentenceSlashNounPhrase = Sentence ** {s2 : Str ; c : Case} ; + + slashTransVerb : Bool -> NounPhrase -> TransVerb -> SentenceSlashNounPhrase = + \b,jussi,ostaa -> + predVerbPhrase jussi (predVerb b ostaa) ** { + s2 = ostaa.s3 ++ ostaa.s4 ; + c = npForm2Case jussi.n + (complementCase b ostaa.c (Ind jussi.n (np2Person jussi.p))) + } ; + +--2 Relative pronouns and relative clauses +-- +-- As described in $types.Fin.gf$, relative pronouns are inflected like +-- common nouns, in number and case. +-- +-- We get the simple relative pronoun "joka" from $morpho.Fin.gf$. + + identRelPron : RelPron = relPron ; + + funRelPron : Function -> RelPron -> RelPron = \vaimo, joka -> + {s = \\n,c => joka.s ! n ! npForm2Case n vaimo.c ++ vaimo.s ! False ! n ! c} ; + +-- Relative clauses can be formed from both verb phrases ("joka ui") and +-- slash expressions ("jonka sinä näet", "jonka kautta sinä käyt"). + + RelClause : Type = {s : Number => Str} ; + + relVerbPhrase : RelPron -> VerbPhrase -> RelClause = \joka,ui -> + {s = \\n => joka.s ! n ! Nom ++ ui.s ! Ind n P3 ++ ui.s2 ! Ind n P3} ; + + relSlash : RelPron -> SentenceSlashNounPhrase -> RelClause = \joka,saat -> + {s = \\n => joka.s ! n ! saat.c ++ saat.s2 ++ saat.s} ; + +-- A 'degenerate' relative clause is the one often used in mathematics, e.g. +-- "luku x siten että x on parillinen". + + relSuch : Sentence -> RelClause = \A -> + {s = \\_ => advSiten ++ conjEtta ++ A.s} ; + +-- N.B. the construction "sellainen että" is not possible with the present +-- typing of the relative clause, since it should also be inflected in +-- case. Ordinary relative clauses have a fixed case. +-- +-- The main use of relative clauses is to modify common nouns. +-- The result is a common noun, out of which noun phrases can be formed +-- by determiners. We use no comma before these relative clauses, even though +-- conservative standard Finnish does. + + modRelClause : CommNounPhrase -> RelClause -> CommNounPhrase = \mies,jokaui -> + {s = \\b,n,c => mies.s ! b ! n ! c ++ jokaui.s ! n ; + g = mies.g + } ; + +-- N.B: the possessive suffix, if attached here, comes to wrong place! Solution: +-- make $CommNounPhrase$ discontinuos. + + +--2 Interrogative pronouns +-- +-- If relative pronouns are like common nouns (and adjectives), +-- interrogative pronouns are like noun phrases, having a fixed number. +-- They also need to handle an NP-like accusative case. But person is +-- not needed, since it is uniformly $NP3$. + + IntPron : Type = {s : NPForm => Str ; n : Number} ; + +-- In analogy with relative pronouns, we have a rule for applying a function +-- to a relative pronoun to create a new one. + + funIntPron : Function -> IntPron -> IntPron = \vaimo,kuka -> + {s = \\c => kuka.s ! vaimo.c ++ + vaimo.s ! False ! kuka.n ! npForm2Case kuka.n c ; + n = kuka.n + } ; + +-- There is a variety of simple interrogative pronouns: +-- "mikä talo" / "kuka mies", "kuka", "mikä". The construction with a noun +-- is the reason why nouns in Finnish need a gender. + + nounIntPron : Number -> CommNounPhrase -> IntPron = \n, talo -> + {s = \\c => let {nc = npForm2Case n c} in + mikakukaInt ! talo.g ! n ! nc ++ talo.s ! False ! n ! nc ; + n = n + } ; + + intPronWho : Number -> IntPron = \num -> { + s = \\c => mikakukaInt ! Human ! num ! (npForm2Case num c) ; + n = num + } ; + + intPronWhat : Number -> IntPron = \num -> { + s = \\c => mikakukaInt ! NonHuman ! num ! (npForm2Case num c) ; + n = num + } ; + + +--2 Utterances + +-- By utterances we mean complete phrases, such as +-- 'can be used as moves in a language game': indicatives, questions, imperative, +-- and one-word utterances. The rules are far from complete. +-- +-- N.B. we have not included rules for texts, which we find we cannot say much +-- about on this level. In semantically rich GF grammars, texts, dialogues, etc, +-- will of course play an important role as categories not reducible to utterances. +-- An example is proof texts, whose semantics show a dependence between premises +-- and conclusions. Another example is intersentential anaphora. + + Utterance = SS ; + + indicUtt : Sentence -> Utterance = \x -> ss (x.s ++ stopPunct) ; + interrogUtt : Question -> Utterance = \x -> ss (x.s ++ questPunct) ; + + +--2 Questions +-- +-- Questions are either direct or indirect, but the forms in Finnish are +-- always identical. So we don't need a $QuestForm$ parameter as in other languages. + +oper + Question = SS ; + +--3 Yes-no questions +-- +-- Yes-no questions are formed by inversed predication, with the clitic "ko" / "kö" +-- particle attached to the verb part of the verb phrase. + + questVerbPhrase : NounPhrase -> VerbPhrase -> Question = \jussi,ui -> + let {np = Ind jussi.n (np2Person jussi.p)} in + ss (ui.s ! np ++ koPart ++ jussi.s ! NPCase Nom ++ ui.s2 ! np) ; + + +--3 Wh-questions +-- +-- Wh-questions are of two kinds: ones that are like $NP - VP$ sentences +-- ("kuka ui?") others that are line $S/NP - NP$ sentences ("kenet sinä tapaat?"). + + intVerbPhrase : IntPron -> VerbPhrase -> Question = \kuka,ui -> + predVerbPhrase (kuka ** {p = NP3}) ui ; + + intSlash : IntPron -> SentenceSlashNounPhrase -> Question = \kuka,tapaat -> + ss (kuka.s ! NPCase tapaat.c ++ tapaat.s2 ++ tapaat.s) ; + + +--3 Interrogative adverbials +-- +-- These adverbials will be defined in the lexicon: they include +-- "koska", "missä", "kuinka", "miksi", etc, which are all invariant one-word +-- expressions. In addition, they can be formed by adding cases and postpositions +-- to interrogative pronouns, in the same way as adverbials are formed +-- from noun phrases; notice that even prepositions are used as postpositions +-- when attached to interrogative pronouns. + + IntAdverb = SS ; + + prepIntAdverb : Str -> Case -> IntPron -> IntAdverb = \ennen,c,kuka -> + ss (kuka.s ! NPCase c ++ ennen) ; + +-- A question adverbial can be applied to anything, and whether this makes +-- sense is a semantic question. The syntax is very simple: just prefix the +-- adverbial to the predication. + + questAdverbial : IntAdverb -> NounPhrase -> VerbPhrase -> Question = + \miksi, jussi, ui -> + cc2 miksi (predVerbPhrase jussi ui) ; + +--2 Imperatives +-- +-- We only consider second-person imperatives. + + Imperative = SS1 Number ; + + imperVerbPhrase : VerbPhrase -> Imperative = \ui -> + {s = \\n => ui.s ! Imper n ++ ui.s2 ! Imper n} ; + + imperUtterance : Number -> Imperative -> Utterance = \n,I -> + ss (I.s ! n ++ exclPunct) ; + +--2 Sentence adverbials +-- +-- This class covers adverbials such as "muuten", "siksi", which are prefixed +-- to a sentence to form a phrase. + + advSentence : Adverb -> Sentence -> Utterance = \siksi,sataa -> + ss (siksi.s ++ sataa.s ++ ".") ; + + +--2 Coordination +-- +-- Coordination is to some extent orthogonal to the rest of syntax, and +-- has been treated in a generic way in the module $CO$ in the file +-- $coordination.gf$. The overall structure is independent of category, +-- but there can be differences in parameter dependencies. +-- +--3 Conjunctions +-- +-- Coordinated phrases are built by using conjunctions, which are either +-- simple ("ja", "tai") or distributed ("sekä - että", "joko - tai"). +-- +-- The conjunction has an inherent number, which is used when conjoining +-- noun phrases: "Jussi ja Mari ovat..." vs. "Jussi tai Mari on..."; in the +-- case of "tai", the result is however plural if any of the disjuncts is. + + Conjunction = CO.Conjunction ** {n : Number} ; + ConjunctionDistr = CO.ConjunctionDistr ** {n : Number} ; + +--3 Coordinating sentences +-- +-- We need a category of lists of sentences. It is a discontinuous +-- category, the parts corresponding to 'init' and 'last' segments +-- (rather than 'head' and 'tail', because we have to keep track of the slot between +-- the last two elements of the list). A list has at least two elements. + + ListSentence : Type = SD2 ; + + twoSentence : (_,_ : Sentence) -> ListSentence = CO.twoSS ; + + consSentence : ListSentence -> Sentence -> ListSentence = + CO.consSS CO.comma ; + +-- To coordinate a list of sentences by a simple conjunction, we place +-- it between the last two elements; commas are put in the other slots, +-- e.g. "du rauchst, er trinkt und ich esse". + + conjunctSentence : Conjunction -> ListSentence -> Sentence = \c,xs -> + ss (CO.conjunctX c xs) ; + +-- To coordinate a list of sentences by a distributed conjunction, we place +-- the first part (e.g. "either") in front of the first element, the second +-- part ("or") between the last two elements, and commas in the other slots. +-- For sentences this is really not used. + + conjunctDistrSentence : ConjunctionDistr -> ListSentence -> Sentence = + \c,xs -> + ss (CO.conjunctDistrX c xs) ; + + +--3 Coordinating adjective phrases +-- +-- The structure is the same as for sentences. Parameters are passed to components. + + ListAdjPhrase : Type = + {s1,s2 : AdjPos => Number => Case => Str} ; + + twoAdjPhrase : (_,_ : AdjPhrase) -> ListAdjPhrase = \x,y -> + CO.twoTable3 AdjPos Number Case x y ; + + consAdjPhrase : ListAdjPhrase -> AdjPhrase -> ListAdjPhrase = \xs,x -> + CO.consTable3 AdjPos Number Case CO.comma xs x ; + + conjunctAdjPhrase : Conjunction -> ListAdjPhrase -> AdjPhrase = \c,xs -> + CO.conjunctTable3 AdjPos Number Case c xs ; + + conjunctDistrAdjPhrase : ConjunctionDistr -> ListAdjPhrase -> AdjPhrase = \c,xs -> + CO.conjunctDistrTable3 AdjPos Number Case c xs ; + + +--3 Coordinating noun phrases +-- +-- The structure is the same as for sentences. The result is either always plural +-- or plural if any of the components is, depending on the conjunction. + + ListNounPhrase : Type = {s1,s2 : NPForm => Str ; n : Number ; p : NPPerson} ; + + twoNounPhrase : (_,_ : NounPhrase) -> ListNounPhrase = \x,y -> + CO.twoTable NPForm x y ** {n = conjNumber x.n y.n ; p = conjPerson x.p y.p} ; + + consNounPhrase : ListNounPhrase -> NounPhrase -> ListNounPhrase = \xs,x -> + CO.consTable NPForm CO.comma xs x ** + {n = conjNumber xs.n x.n ; p = conjPerson xs.p x.p} ; + + conjunctNounPhrase : Conjunction -> ListNounPhrase -> NounPhrase = \c,xs -> + CO.conjunctTable NPForm c xs ** {n = conjNumber c.n xs.n ; p = xs.p} ; + + conjunctDistrNounPhrase : ConjunctionDistr -> ListNounPhrase -> NounPhrase = + \c,xs -> + CO.conjunctDistrTable NPForm c xs ** {n = conjNumber c.n xs.n ; p = xs.p} ; + +-- We have to define a calculus of numbers of persons. For numbers, +-- it is like the conjunction with $Pl$ corresponding to $False$. + + conjNumber : Number -> Number -> Number = \m,n -> case of { + => Sg ; + _ => Pl + } ; + +-- For persons, we let the latter argument win ("either you or I am absent" +-- but "either I or you are absent"). This is not quite clear. + + conjPerson : NPPerson -> NPPerson -> NPPerson = \_,p -> + p ; + + + +--2 Subjunction +-- +-- Subjunctions ("kun", "jos", etc) +-- are a different way to combine sentences than conjunctions. +-- The main clause can be a sentences, an imperatives, or a question, +-- but the subjoined clause must be a sentence. +-- +-- There are uniformly two variant word orders, e.g. +-- "jos poltat minä suutun" +-- and "minä suutun jos poltat". + + Subjunction = SS ; + + subjunctSentence : Subjunction -> Sentence -> Sentence -> Sentence = + \if, A, B -> + ss (subjunctVariants if A.s B.s) ; + + subjunctImperative : Subjunction -> Sentence -> Imperative -> Imperative = + \if, A, B -> + {s = \\n => subjunctVariants if A.s (B.s ! n)} ; + + subjunctQuestion : Subjunction -> Sentence -> Question -> Question = + \if, A, B -> + {s = subjunctVariants if A.s B.s} ; + + subjunctVariants : Subjunction -> Str -> Str -> Str = \if,A,B -> + variants {if.s ++ A ++ commaPunct ++ B ; B ++ commaPunct ++ if.s ++ A} ; + +--2 One-word utterances +-- +-- An utterance can consist of one phrase of almost any category, +-- the limiting case being one-word utterances. These +-- utterances are often (but not always) in what can be called the +-- default form of a category, e.g. the nominative. +-- This list is far from exhaustive. + + useNounPhrase : NounPhrase -> Utterance = \john -> + postfixSS stopPunct (defaultNounPhrase john) ; + + useCommonNounPhrase : Number -> CommNounPhrase -> Utterance = \n,car -> + useNounPhrase (indefNounPhrase n car) ; + +-- Here are some default forms. + + defaultNounPhrase : NounPhrase -> SS = \john -> + ss (john.s ! NPCase Nom) ; + + defaultQuestion : Question -> SS = \whoareyou -> + whoareyou ; + + defaultSentence : Sentence -> Utterance = \x -> + x ; +} ; diff --git a/grammars/resource/finnish/TestFin.gf b/grammars/resource/finnish/TestFin.gf new file mode 100644 index 000000000..79ea285af --- /dev/null +++ b/grammars/resource/finnish/TestFin.gf @@ -0,0 +1,40 @@ +concrete TestFin of TestAbs = ResFin ** open Prelude, Syntax in { + +flags startcat=Phr ; lexer=text ; parser=chart ; unlexer=text ; + +-- a random sample from the lexicon + +lin + Big = regAdjDegr (sTalo "iso") "isompaa" "isointa" ; + Small = regAdjDegr (sSusi "pieni" "pienen" "pienenä") "pienempää" "pienintä" ; + Old = regAdjDegr (sKukko "vanha" "vanhan" "vanhoja") "vanhempaa" "vanhinta" ; + Young = regAdjDegr (sSusi "nuori" "nuoren" "nuorena") "nuorempaa" "nuorinta" ; + + Man = cnHum (mkNoun "mies" "miehen" "miehenä" "miestä" "mieheen" "miehinä" + "miehissä" "miesten" "miehiä" "miehiin") ; + Woman = cnHum (sNainen "naista") ; + Car = cnNoHum (sTalo "auto") ; + House = cnNoHum (sTalo "talo") ; + Light = cnNoHum (sTalo "valo") ; + + Walk = vJuosta "kävellä" "kävelen" ; + Run = vJuosta "juosta" "juoksen" ; + Say = vSanoa "sanoa" ; + Prove = vPoistaa "todistaa" ; + Send = mkTransVerbDir (vOttaa "lähettää" "lähetän") ; + Love = mkTransVerbCase (vPoistaa "rakastaa") Part ; + Wait = mkTransVerbCase (vOttaa "odottaa" "odotan") Part ; + + Mother = funGen (n2n (cnHum (sKukko "äiti" "äidin" "äitejä"))) ; + Uncle = funGen (n2n (cnHum (sKukko "setä" "sedän" "setiä"))) ; --- eno! + + Always = ss "aina" ; + Well = ss "hyvin" ; + + SwitchOn = mkTransVerbDir (vOttaa "sytyttää" "sytytän") ; + SwitchOff = mkTransVerbDir (vOttaa "sammuttaa" "sammutan") ; + + John = mkProperName (sKukko "Jussi" "Jussin" "Jusseja") ; + Mary = mkProperName (sKukko "Mari" "Marin" "Mareja") ; + +} ; \ No newline at end of file diff --git a/grammars/resource/finnish/Types.gf b/grammars/resource/finnish/Types.gf new file mode 100644 index 000000000..9c2193d00 --- /dev/null +++ b/grammars/resource/finnish/Types.gf @@ -0,0 +1,123 @@ +--1 Finnish Word Classes and Morphological Parameters +-- +-- This is a resource module for Finnish morphology, defining the +-- morphological parameters and word classes of Finnish. It is aimed +-- to be complete w.r.t. the description of word forms. +-- However, it only includes those parameters that are needed for +-- analysing individual words: such parameters are defined in syntax modules. +-- +-- We use the language-independent prelude. + +resource Types = open Prelude in { + +-- +--2 Enumerated parameter types +-- +-- These types are the ones found in school grammars. +-- Their parameter values are atomic. We omit three of the cases, not +-- occurring in the resource syntax. The accusative cases are only +-- defined in syntax; in morphology, there is a special accusative for +-- pronouns. + +param + Number = Sg | Pl ; + Case = Nom | Gen | Part | Transl | Ess + | Iness | Elat | Illat | Adess | Ablat | Allat ; + Person = P1 | P2 | P3 ; + Degree = Pos | Comp | Sup ; + Gender = NonHuman | Human ; + +-- For data abstraction, we define + +oper + singular = Sg ; + plural = Pl ; + +--2 Word classes and hierarchical parameter types +-- +-- Real parameter types (i.e. ones on which words and phrases depend) +-- are often hierarchical. The alternative would be cross-products of +-- simple parameters, but this would usually overgenerate. +-- + +--3 Common nouns +-- +-- Common nouns are inflected in number and noun case. In noun case, we include +-- forms used in connection with possessive suffixes. + +param + NForm = NCase Number Case + | NPossNom | NPossGenPl | NPossTransl Number | NPossIllat Number ; + +oper + CommonNoun : Type = {s : NForm => Str} ; + + useNForm : NForm -> (Number => Case => Str) -> Str = \nf,f -> case nf of { + NCase n c => f ! n ! c ; + NPossNom => f ! Sg ! Nom ; ---- "iso autoni"; also "isot autoni" etc + NPossGenPl => f ! Pl ! Gen ; + NPossTransl n => f ! n ! Transl ; + NPossIllat n => f ! n ! Illat + } ; + + +-- +--3 Adjectives +-- +-- The major division is between the comparison degrees, but it +-- is also good to leave room for adjectives that cannon be compared. +-- Such adjectives are like common nouns. + + Adjective : Type = CommonNoun ; + AdjDegr : Type = {s : Degree => NForm => Str} ; + +--3 Verbs +-- +-- We limit the grammar so far to verbs in the infinitive, second-person +-- imperative, and present tense indicative. A special form is needed for +-- the negated plural imperative. + +param + VForm = + Inf + | Ind Number Person + | Imper Number + | ImpNegPl ; + +oper + Verb : Type = SS1 VForm ; + + vFormNeg = Imper Sg ; + + vform2number : VForm -> Number = \v -> case v of { + Inf => Sg ; + Ind n _ => n ; + Imper n => n ; + ImpNegPl => Pl + } ; + +-- +--3 Pronouns +-- +-- For pronouns, we need the noun case forms, plus an accusative. + +param + PForm = PCase Case | PAcc ; + +oper + Pronoun : Type = {s : PForm => Str ; n : Number ; p : Person} ; + +--3 Proper names +-- +-- Proper names only need case forms. + + ProperName : Type = SS1 Case ; + + +--3 Relative pronouns +-- +-- Relative pronouns are inflected like nouns, except for possessive suffixes. + + RelPron : Type = {s : Number => Case => Str} ; + +} ; \ No newline at end of file diff --git a/src/GF/Canon/Look.hs b/src/GF/Canon/Look.hs index ec76008f9..4318239b6 100644 --- a/src/GF/Canon/Look.hs +++ b/src/GF/Canon/Look.hs @@ -112,6 +112,9 @@ ccompute cnc = comp [] if noVar v' then matchPatt cs v' >>= compt else return $ S u' v' + FV ccs -> do + v' <- compt v + mapM (\c -> compt (S c v')) ccs >>= return . FV _ -> liftM (S u') $ compt v diff --git a/src/GF/Compile/CheckGrammar.hs b/src/GF/Compile/CheckGrammar.hs index 07151d8b7..8fe4cf988 100644 --- a/src/GF/Compile/CheckGrammar.hs +++ b/src/GF/Compile/CheckGrammar.hs @@ -101,7 +101,7 @@ checkResInfo gr (c,info) = do ty' <- check ty typeType >>= comp . fst (de',_) <- check de ty' return (Yes ty', Yes de') - (Nope, Yes de) -> do + (_, Yes de) -> do (de',ty') <- infer de return (Yes ty', Yes de') _ -> return (pty, pde) --- other cases are uninteresting @@ -611,6 +611,10 @@ checkEqLType env t u trm = do || elem n (allExtends env m) (QC m a, QC n b) | a == b -> elem m (allExtends env n) || elem n (allExtends env m) + (QC m a, Q n b) | a == b -> elem m (allExtends env n) + || elem n (allExtends env m) + (Q m a, QC n b) | a == b -> elem m (allExtends env n) + || elem n (allExtends env m) (RecType rs, RecType ts) -> and [alpha g a b && l == k --- too strong req | ((l,a),(k,b)) <- zip rs ts] diff --git a/src/GF/Compile/Extend.hs b/src/GF/Compile/Extend.hs index 582a1e6ae..348cdf71d 100644 --- a/src/GF/Compile/Extend.hs +++ b/src/GF/Compile/Extend.hs @@ -27,8 +27,9 @@ extendModInfo name old new = case (old,new) of extendMod :: Ident -> BinTree (Ident,Info) -> BinTree (Ident,Info) -> Err (BinTree (Ident,Info)) -extendMod name old new = - foldM (tryInsert (extendAnyInfo name) (indirInfo name)) new $ tree2list old +extendMod name old new = foldM try new $ tree2list old where + try t i@(c,_) = errIn ("constant" +++ prt c) $ + tryInsert (extendAnyInfo name) (indirInfo name) t i indirInfo :: Ident -> Info -> Info indirInfo n info = AnyInd b n' where @@ -58,7 +59,7 @@ perhIndir n p = case p of _ -> p extendAnyInfo :: Ident -> Info -> Info -> Err Info -extendAnyInfo n i j = case (i,j) of +extendAnyInfo n i j = errIn ("building extension for" +++ prt n) $ case (i,j) of (AbsCat mc1 mf1, AbsCat mc2 mf2) -> liftM2 AbsCat (updatePerhaps n mc1 mc2) (updatePerhaps n mf1 mf2) --- add cstrs (AbsFun mt1 md1, AbsFun mt2 md2) -> @@ -66,8 +67,7 @@ extendAnyInfo n i j = case (i,j) of (ResParam mt1, ResParam mt2) -> liftM ResParam $ updatePerhaps n mt1 mt2 (ResValue mt1, ResValue mt2) -> liftM ResValue $ updatePerhaps n mt1 mt2 - (ResOper mt1 m1, ResOper mt2 m2) -> - liftM2 ResOper (updatePerhaps n mt1 mt2) (updatePerhaps n m1 m2) + (ResOper mt1 m1, ResOper mt2 m2) -> extendResOper n mt1 m1 mt2 m2 (CncCat mc1 mf1 mp1, CncCat mc2 mf2 mp2) -> liftM3 CncCat (updatePerhaps n mc1 mc2) @@ -75,4 +75,20 @@ extendAnyInfo n i j = case (i,j) of (CncFun m mt1 md1, CncFun _ mt2 md2) -> liftM2 (CncFun m) (updatePerhaps n mt1 mt2) (updatePerhaps n md1 md2) - _ -> Bad $ "cannot unify information for" +++ show n + (AnyInd _ _, ResOper _ _) -> return j ---- + + _ -> Bad $ "cannot unify information in" ++++ show i ++++ "and" ++++ show j + + +-- opers declared in one module and defined in an extension are a special case + +extendResOper n mt1 m1 mt2 m2 = case (m1,m2) of + (Nope,_) -> return $ ResOper (strip mt1) m2 + _ -> liftM2 ResOper (updatePerhaps n mt1 mt2) (updatePerhaps n m1 m2) + where + strip (Yes t) = Yes $ strp t + strip m = m + strp t = case t of + Q _ c -> Vr c + QC _ c -> Vr c + _ -> composSafeOp strp t diff --git a/src/GF/Compile/Rename.hs b/src/GF/Compile/Rename.hs index a4d9b9365..6f652820a 100644 --- a/src/GF/Compile/Rename.hs +++ b/src/GF/Compile/Rename.hs @@ -62,7 +62,7 @@ extendModule ms (name,mod) = case mod of _ -> Bad $ "cannot find extended module" +++ prt n extendMod n (jments m0) js _ -> return js - return $ (name,ModMod (Module mt fs Nothing ops js1)) + return $ (name,ModMod (Module mt fs me ops js1)) type Status = (StatusTree, [(OpenSpec Ident, StatusTree)]) @@ -72,7 +72,9 @@ type StatusTree = BinTree (Ident,StatusInfo) type StatusInfo = Ident -> Term renameIdentTerm :: Status -> Term -> Err Term -renameIdentTerm env@(act,imps) t = case t of +renameIdentTerm env@(act,imps) t = + errIn ("atomic term" +++ prt t +++ "given" +++ unwords (map (prt . fst) qualifs)) $ + case t of Vr c -> do f <- lookupTreeMany prt opens c return $ f c @@ -90,7 +92,8 @@ renameIdentTerm env@(act,imps) t = case t of _ -> return t where opens = act : [st | (OSimple _,st) <- imps] - qualifs = [ (m, st) | (OQualif m _, st) <- imps] + qualifs = [(m, st) | (OQualif m _, st) <- imps] ++ + [(m, st) | (OSimple m, st) <- imps] -- qualifying is always possible --- would it make sense to optimize this by inlining? renameIdentPatt :: Status -> Patt -> Err Patt @@ -117,8 +120,9 @@ tree2status o = case o of buildStatus :: SourceGrammar -> Ident -> SourceModInfo -> Err Status buildStatus gr c mo = let mo' = self2status c mo in case mo of ModMod m -> do - let ops = allOpens m - mods <- mapM (lookupModule gr . openedModule) ops + let gr1 = MGrammar $ (c,mo) : modules gr + ops = [OSimple e | e <- allExtends gr1 c] ++ allOpens m + mods <- mapM (lookupModule gr1 . openedModule) ops let sts = map modInfo2status $ zip ops mods return $ if isModCnc m then (NT, sts) -- the module itself does not define any names diff --git a/src/GF/Data/Operations.hs b/src/GF/Data/Operations.hs index 6dd78e1bc..d4c13f1f4 100644 --- a/src/GF/Data/Operations.hs +++ b/src/GF/Data/Operations.hs @@ -168,6 +168,14 @@ updatePerhaps old p1 p2 = case (p1,p2) of (_, May a) -> Bad "strange indirection" _ -> unifPerhaps p1 p2 +-- here the value is copied instead of referred to; used for oper types +updatePerhapsHard :: b -> Perhaps a b -> Perhaps a b -> Err (Perhaps a b) +updatePerhapsHard old p1 p2 = case (p1,p2) of + (Yes a, Nope) -> return $ yes a + (May older,Nope) -> return $ may older + (_, May a) -> Bad "strange indirection" + _ -> unifPerhaps p1 p2 + -- binary search trees data BinTree a = NT | BT a (BinTree a) (BinTree a) deriving (Show,Read) diff --git a/src/GF/Grammar/Lookup.hs b/src/GF/Grammar/Lookup.hs index b8afbc21e..83d6787ef 100644 --- a/src/GF/Grammar/Lookup.hs +++ b/src/GF/Grammar/Lookup.hs @@ -17,6 +17,7 @@ lookupResDef gr m c = do info <- lookupInfo mo c case info of ResOper _ (Yes t) -> return $ qualifAnnot m t + ResOper _ Nope -> return $ Q m c AnyInd _ n -> lookupResDef gr n c ResParam _ -> return $ QC m c ResValue _ -> return $ QC m c @@ -31,6 +32,7 @@ lookupResType gr m c = do info <- lookupInfo mo c case info of ResOper (Yes t) _ -> return $ qualifAnnot m t + ResOper (May n) _ -> lookupResType gr n c AnyInd _ n -> lookupResType gr n c ResParam _ -> return $ typePType ResValue (Yes t) -> return $ qualifAnnotPar m t diff --git a/src/Today.hs b/src/Today.hs index 8663db727..4c0ebb181 100644 --- a/src/Today.hs +++ b/src/Today.hs @@ -1 +1 @@ -module Today where today = "Fri Oct 10 14:16:56 CEST 2003" +module Today where today = "Tue Oct 21 17:20:02 CEST 2003"