Merge pull request #397 from inariksit/estonian-interrogative-agreement

Estonian interrogative agreement

src/estonian/CatEst.gf

@@ -22,7 +22,7 @@ concrete CatEst of Cat = CommonX ** open HjkEst, ResEst, Prelude in {
     QCl    = {s : ResEst.Tense => Anteriority => Polarity => Str} ;
     IP     = ResEst.IPhrase ;
     IComp  = {s : Agr => Str} ;
-    IDet   = {s : Case => Str ; n : Number ; isNum : Bool} ;
+    IDet   = ResEst.IDeterminer ;
     IQuant = {s : Number => Case => Str} ;
 
 -- Relative
@@ -98,8 +98,8 @@ concrete CatEst of Cat = CommonX ** open HjkEst, ResEst, Prelude in {
     NP = linNP (NPCase Nom) ;
     CN = linCN (NCase Sg Nom) ;
     V,VS,VQ = linV ;
-    V2,VA,V2S,V2Q = linV2 ;
+    V2,VA,V2S,V2Q,V3 = linV2 ;
-
+    IDet = linIDet ;
 
 
 }

src/estonian/LexiconEst.gf

@@ -160,7 +160,7 @@ lin
   science_N = mkN "teadus" ;
   sea_N = mkN "meri" "mere" "merd" "merre" "merede" "meresid" ;
   seek_V2 = mkV2 (mkV "otsima") cpartitive ;
-  see_V2 = mkV2 (mkV "nägema" "näha") ;
+  see_V2 = mkV2 (mkV "nägema" "näha") cpartitive ;
   sell_V3 = mkV3 (mkV "müüma" "müüa" "müüb" "müüakse" "müüge" "müüs" "müünud" "müüdud") accPrep callative ;
   send_V3 = mkV3 (mkV "saatma") accPrep callative ;
   sheep_N = mkN "lammas" "lamba" "lammast" ;

src/estonian/QuestionEst.gf

@@ -44,26 +44,27 @@ concrete QuestionEst of Question = CatEst ** open ResEst, Prelude in {
       postmod = ip.postmod ++ adv.s ;
       } ;
 
--- The computation of $ncase$ is a special case of that in $NounEst.DetCN$,
+-- The computation of $IdetCN$ is a special case of that in $NounEst.DetCN$,
--- since we don't have possessive suffixes or definiteness.
+-- because the interrogative doesn't agree.
---- It could still be nice to have a common oper...
 
-    IdetCN idet cn = let n = idet.n in emptyIP ** {
+   IdetCN idet cn = emptyIP ** {
       s = \\c =>
         let
           k : Case = npform2case n c ;
-          icase : Case = Nom ; --case k of {  --mis kassiga
+          ncase : NForm = case <k,idet.isNum> of {
-         --   (Ess|Abess|Comit|Termin) => Gen ;
+            <Nom,  True> => NCase Sg Part ; -- TODO estonian example (Fin was "mitkä kolme kytkintä")
-         --   _  => k
+            <_,    True> => NCase Sg k ;    -- TODO estonian example (Fin was "miksi kolmeksi kytkimeksi")
-         -- } ;
+            _            => NCase n  k      -- TODO estonian example (Fin was "mitkä kytkimet")
-          ncase : NForm = case <icase,idet.isNum> of {
-            <Nom,  True> => NCase Sg Part ; -- mitkä kolme kytkintä
-            <_,    True> => NCase Sg k ;    -- miksi kolmeksi kytkimeksi
-            _            => NCase n  k      -- mitkä kytkimet
           }
         in
-        idet.s ! icase ++ cn.s ! ncase ;
+          idet.s ! Nom ++        -- mis
-      n = n
+          idet.post ! k ++       -- kolme
+          cn.s ! ncase ;         -- kassi+ga
+      n = idet.n ; -- needed for agreement, "mis kolm kassi mängivad"
+  } where {
+    n : Number = case idet.isNum of {
+                  True  => Sg ;
+                  False => idet.n } ;
   } ;
 
     IdetIP idet = let n = idet.n in emptyIP ** {
@@ -71,26 +72,20 @@ concrete QuestionEst of Question = CatEst ** open ResEst, Prelude in {
         let
           k = npform2case n c ;
         in
-        idet.s ! k ;
+        case idet.isNum of {
-      n = n
+          True  => idet.s ! Nom ++ idet.post ! k ;
+          False => idet.s ! k ++ idet.post ! k
         } ;
 
-    IdetQuant idet num =
+      n = n
-      let
+      } ;
-        n = num.n ;
+    -- The quant and the num may be inflected in different cases:
-        isn = num.isNum
+      -- * mis kolme koeraga, mis kolmega
-      in {
+      -- * millega
-        s = \\k =>
+      -- * mille 3-ga (this would be the preferable output, but currently outputs "mis 3ga")
-        let
+    IdetQuant idet num = num ** {
-          ncase = case <k,isn> of {
+      s = \\c => idet.s ! num.n ! c ;
-            <Nom,  True> => NCase Sg Part ; -- mitkä kolme kytkintä
+      post = \\c => num.s ! Sg ! c ;
-            <_,    True> => NCase Sg k ;    -- miksi kolmeksi kytkimeksi
-            _            => NCase n  k      -- mitkä kytkimet
-            }
-        in
-        idet.s ! n ! k ++ num.s ! Sg ! k ;
-      n = n ;
-      isNum = isn
       } ;
 
     AdvIAdv i a = {s = i.s ++ a.s} ;

src/estonian/ResEst.gf

@@ -806,7 +806,9 @@ oper
     isDef : Bool             -- True (verb agrees in Pl, Nom is not Part) --I: actually, can we get rid of this?
     } ;
 
-  IDeterminer : Type = {s : Case => Str ; n : Number ; isNum : Bool} ;
+  IDeterminer : Type = {s, post : Case => Str ; n : Number ; isNum : Bool} ;
+
+  linIDet : IDeterminer -> Str = \idet -> idet.s ! Nom ++ idet.post ! Nom ;
 
     heavyDet : PDet -> Determiner = \d -> d ** {sp = d.s} ;
     PDet : Type = {

src/estonian/StructuralEst.gf

@@ -48,6 +48,7 @@ concrete StructuralEst of Structural = CatEst **
   how8much_IAdv = ss "kui palju" ;
   how8many_IDet = {
     s = \\c => "kui" ++ (mkN "mitu" "mitme" "mitut" "TODO" "TODO" "TODO").s ! NCase Sg c ;
+    post = \\c => [] ;
     n = Sg ;
     isNum = False
   } ;

src/estonian/unittest/comitative.gftest

@@ -12,11 +12,11 @@
 
 Lang: PredVP (UsePron i_Pron) (AdvVP (UseV play_V) (PrepNP with_Prep (DetCN (DetQuant this_Quant NumSg) (UseN dog_N))))
 LangEng: I play with this dog
-LangEst: mina mängin koos selle koera &+ ga
+LangEst: mina mängin koos selle koera BIND ga
 
 Lang: PredVP (UsePron we_Pron) (AdvVP (UseV play_V) (PrepNP with_Prep (DetCN (DetQuant this_Quant NumPl) (UseN dog_N))))
 LangEng: we play with these dogs
-LangEst: meie mängime koos nende koerade &+ ga
+LangEst: meie mängime koos nende koerte BIND ga
 
 -- Interrogative versions
 
@@ -33,37 +33,32 @@ LangEst: mis keelt sina räägid
 
 Lang: QuestIAdv (PrepIP with_Prep (IdetCN (IdetQuant which_IQuant NumSg) (UseN dog_N))) (PredVP (UsePron youSg_Pron) (UseV play_V))
 LangEng: with which dog do you play
-LangEst: koos mis koera &+ ga sina mängid
+LangEst: koos mis koera BIND ga sina mängid
 
 Lang: QuestIAdv (PrepIP with_Prep (IdetCN (IdetQuant which_IQuant NumPl) (UseN dog_N))) (PredVP (UsePron youPl_Pron) (UseV play_V))
 LangEng: with which dogs do you play
-LangEst: koos mis koerade &+ ga teie mängite
+LangEst: koos mis koerte BIND ga teie mängite
 
 -- Adding a number, does it affect the behaviour of
 -- a) suffix attachment? b) agreement of interrogative?
 
 Lang: PredVP (UsePron i_Pron) (AdvVP (UseV play_V) (PrepNP with_Prep (DetCN (DetQuant this_Quant (NumCard (NumNumeral (num (pot2as3 (pot1as2 (pot0as1 (pot0 n3)))))))) (UseN dog_N))))
 LangEng: I play with these three dogs
-LangEst: mina mängin koos nende kolme koera &+ ga
+LangEst: mina mängin koos nende kolme koera BIND ga
 
 Lang: PredVP (UsePron i_Pron) (AdvVP (UseV play_V) (PrepNP with_Prep (DetCN (DetQuant this_Quant (NumCard (NumDigits (IDig D_3)))) (UseN dog_N))))
 LangEng: I play with these 3 dogs
-LangEst: mina mängin koos nende 3 koera &+ ga
+LangEst: mina mängin koos nende 3 koera BIND ga
 
 Lang: QuestIAdv (PrepIP with_Prep (IdetCN (IdetQuant which_IQuant (NumCard (NumNumeral (num (pot2as3 (pot1as2 (pot0as1 (pot0 n3)))))))) (UseN dog_N))) (PredVP (UsePron youPol_Pron) (UseV play_V))
 LangEng: with which three dogs do you play
--- Which one is best?
+LangEst: koos mis kolme koera BIND ga teie mängite
--- 1) koos mis kolm koeraga teie mängite
--- 2) koos mis kolme koeraga teie mängite
--- 3) koos mille kolme koeraga teie mängite
--- 4) All are ungrammatical (provide more grammatical translation that is still compositional, not paraphrase)
 
 Lang: QuestIAdv (PrepIP with_Prep (IdetCN (IdetQuant which_IQuant (NumCard (NumDigits (IDig D_3)))) (UseN dog_N))) (PredVP (UsePron youPl_Pron) (UseV play_V))
 LangEng: with which 3 dogs do you play
--- Which one is best?
+LangEst: koos mis 3 koera BIND ga teie mängite
--- 1) koos mis 3 koeraga teie mängite
+-- Technically the version below is better, but we'd need to add another param and who wants that :-P
--- 2) koos mille 3 koeraga teie mängite
+-- LangEst: koos mille 3 koera BIND ga teie mängite
--- 3) All are ungrammatical (provide more grammatical translation)
 
 ------------------------------------------------------------------------------------------
 -- Now we skip the content word, and make the determiner (possibly with number) into NP/IP
@@ -72,75 +67,60 @@ LangEng: with which 3 dogs do you play
 -- No number, case attaches to determiner/interrogative
 Lang: PredVP (UsePron i_Pron) (AdvVP (UseV play_V) (PrepNP with_Prep (DetNP (DetQuant this_Quant NumPl))))
 LangEng: I play with these
-LangEst: mina mängin koos nende &+ ga
+LangEst: mina mängin koos nende BIND ga
 
 Lang: QuestIAdv (PrepIP with_Prep whoSg_IP) (PredVP (UsePron youSg_Pron) (UseV sing_V))
 LangEng: with whom do you sing
-LangEst: koos kelle &+ ga sina laulad
+LangEst: koos kelle BIND ga sina laulad
 
 Lang: QuestIAdv (PrepIP with_Prep whoPl_IP) (PredVP (UsePron youSg_Pron) (UseV sing_V))
 LangEng: with whom do you sing
-LangEst: koos kellede &+ ga sina laulad
+LangEst: koos kellede BIND ga sina laulad
 
 -- Adding a number to a determiner, does this affect the comitative suffix placement?
 -- Context of the sentence is an ellipsis, "I sing with these 3 (singers)"
 
 Lang: PredVP (UsePron i_Pron) (AdvVP (UseV sing_V) (PrepNP with_Prep (DetNP (DetQuant this_Quant (NumCard (NumNumeral (num (pot2as3 (pot1as2 (pot0as1 (pot0 n3)))))))))))
 LangEng: I sing with these three
--- Which one is better?
+LangEst: mina laulan koos nende kolme BIND ga
--- 1) mina laulan koos nende kolmega
--- 2) mina laulan koos nendega kolme
 
+-- Orthography should be 3-ga, but this can be easily postprocessed in applications, so I don't bother
 Lang: PredVP (UsePron i_Pron) (AdvVP (UseV sing_V) (PrepNP with_Prep (DetNP (DetQuant this_Quant (NumCard (NumDigits (IDig D_3)))))))
 LangEng: I sing with these 3
-LangEst: mina laulan koos nende 3 &+ ga
+LangEst: mina laulan koos nende 3 BIND ga
--- Which one is better?
--- 1) mina laulan koos nende 3ga
--- 2) mina laulan koos nendega 3
 
 -- Adding a number to an interrogative, how does suffix placement + agreement work now?
 
 Lang: QuestIAdv (PrepIP with_Prep (IdetIP (IdetQuant which_IQuant (NumCard (NumNumeral (num (pot2as3 (pot1as2 (pot0as1 (pot0 n3)))))))))) (PredVP (UsePron youSg_Pron) (UseV sing_V))
 LangEng: with which three do you sing
--- Which one is best?
+LangEst: koos mis kolme BIND ga sina laulad
--- 1) koos millede kolmega sina laulad
--- 2) koos mille kolmega sina laulad
--- 3) koos mis kolmega sina laulad
--- 4) koos millega kolme sina laulad
--- 5) koos milledega kolme sina laulad
--- 6) All are ungrammatical (provide more grammatical translation)
 
 Lang: QuestIAdv (PrepIP with_Prep (IdetIP (IdetQuant which_IQuant (NumCard (NumDigits (IDig D_3)))))) (PredVP (UsePron youSg_Pron) (UseV sing_V))
 LangEng: with which 3 do you sing
--- Which one is best?
+LangEst: koos mis 3 BIND ga sina laulad
--- 1) koos millede 3ga sina laulad
+-- Also ok:
--- 2) koos mille 3ga sina laulad
+-- LangEst: koos mille 3 BIND ga sina laulad
--- 3) koos mis 3ga sina laulad
--- 4) koos millega 3 sina laulad
--- 5) koos milledega 3 sina laulad
--- 6) All are ungrammatical (provide more grammatical translation)
-
 
 -- Adding premodifiers to the number, does that change anything?
 
--- Weird sentence, but imagine like, you point to a group of people, "I sing with these at most three",
+-- None of the alternatives sounds actually good, so we just use the same strategy as interrogative + number
--- like you are the soloist and want at most 3 backup singers, who are among a group you are pointing at.
+-- This can be one of the cases where some trees in the RGL don't make sense, and so application grammarians should use other trees that make more sense.
--- If there was no ellipsis, the phrase would just be "koos nende kuni kolme lauljatega"?
+-- Leaving these ungrammatical/nonsensical edge cases here just for documentation, and to catch any unintended side effects of later tweaks to the grammar.
 Lang: PredVP (UsePron i_Pron) (AdvVP (UseV sing_V) (PrepNP with_Prep (DetNP (DetQuant this_Quant (NumCard (AdNum at_most_AdN (NumNumeral (num (pot2as3 (pot1as2 (pot0as1 (pot0 n3))))))))))))
 LangEng: I sing with these at most three
-LangEst: mina laulan koos nende kuni kolme &+ ga
+LangEst: mina laulan koos nende kuni kolme BIND ga
--- Which one is better?
--- 1) mina laulan koos nende kuni kolmega
--- 2) mina laulan koos nendega kuni kolme
 
 Lang: QuestIAdv (PrepIP with_Prep (IdetIP (IdetQuant which_IQuant (NumCard (AdNum at_most_AdN (NumNumeral (num (pot2as3 (pot1as2 (pot0as1 (pot0 n3))))))))))) (PredVP (UsePron youSg_Pron) (UseV sing_V))
 LangEng: with which at most three do you sing
-LangEst: koos millede kuni kolme &+ ga sina laulad
+LangEst: koos mis kuni kolme BIND ga sina laulad
--- Which one is best?
--- 1) koos millede kuni kolmega sina laulad
--- 2) koos mille kuni kolmega sina laulad
--- 3) koos mis kuni kolmega sina laulad
--- 4) koos millega kuni kolme sina laulad
--- 5) koos milledega kuni kolme sina laulad
--- 6) All are ungrammatical (provide more grammatical translation)
 
+-- Appendix: if you really want to ask "with which at most three do you sing", you can use these strategies
+
+-- koos kellega (kuni kolmega) sina laulad? -- this one is by far the best
+-- koos kellega (kuni kolm) sina laulad?    --
+-- koos kellega (kuni 3) sina laulad?       --
+-- koos kellega (kuni 3-ga) sina laulad?    --
+-- Now I get it, the sentence doesn't have any asking verb, "kellega koos sina laulad" VS "kellega koos sina *tahad* laulda?"
+-- If you want the answerer to specify the up-to-three to sing together with, then you have to let specify also whether-or-not the answerer even wants to sing :-)
+-- *) kuni kolmega nendest, kellega sina laulaud? -- this is correct but just weird
+-- *) kuni kolmega nendest, kellega sina tahad laulda? -- this is correct and understandable

src/estonian/unittest/interrogative-agreement.gftest

@@ -0,0 +1,52 @@
+-- Test cases related to the agreement of interrogative pronouns
+-- for tests about interrogative agreement in comitative, see comitative.gftest
+
+-- As object: case of the numberal is determined by the verb's object case
+
+-- For comparison, the Finnish linearisation is
+-- LangFin: mitkä kahdeksan opiskelijaa sinä näet
+-- Estonian partitive here comes from see_V2
+Lang: QuestSlash (IdetCN (IdetQuant which_IQuant (NumCard (NumNumeral (num (pot2as3 (pot1as2 (pot0as1 (pot0 n8)))))))) (UseN student_N)) (SlashVP (UsePron youSg_Pron) (SlashV2a see_V2))
+LangEng: which eight students do you see
+LangEst: mis kaheksat tudengit sina näed
+
+-- In Finnish, partitive comes from the number (mitkä kahdeksan opiskelijaa sinä myyt opettajille)
+-- In Estonian, we have eight and student both in genitive.
+Lang: QuestSlash (IdetCN (IdetQuant which_IQuant (NumCard (NumNumeral (num (pot2as3 (pot1as2 (pot0as1 (pot0 n8)))))))) (UseN student_N)) (SlashVP (UsePron youSg_Pron) (Slash3V3 sell_V3 (DetCN (DetQuant DefArt NumPl) (UseN teacher_N))))
+LangEng: which eight students do you sell to the teachers
+LangEst: mis kaheksa tudengi sina müüd õpetajatele
+
+-- Both Finnish and Estonian inflect eight and student, Finnish also inflects which (mistä kahdeksasta opiskelijasta sinä pidät)
+Lang: QuestSlash (IdetCN (IdetQuant which_IQuant (NumCard (NumNumeral (num (pot2as3 (pot1as2 (pot0as1 (pot0 n8)))))))) (UseN student_N)) (SlashVP (UsePron youSg_Pron) (SlashV2a like_V2))
+LangEng: which eight students do you like
+LangEst: mis kaheksast tudengist sina pead lugu
+
+-- For comparison, this is Finnish partitive coming from the verb: the word "eight" is also inflected
+Lang: QuestSlash (IdetCN (IdetQuant which_IQuant (NumCard (NumNumeral (num (pot2as3 (pot1as2 (pot0as1 (pot0 n8)))))))) (UseN student_N)) (SlashVP (UsePron youSg_Pron) (SlashV2a love_V2))
+LangEng: which eight students do you love
+LangEst: mis kaheksat tudengit sina armastad
+-- LangFin: mitä kahdeksaa opiskelijaa sinä rakastat
+
+
+-- As subject
+
+-- No numeral in IDet
+Lang: QuestVP (IdetCN (IdetQuant which_IQuant NumSg) (UseN horse_N)) (UseV play_V)
+LangEng: which horse plays
+LangEst: mis hobune mängib
+
+Lang: QuestVP (IdetCN (IdetQuant which_IQuant NumPl) (UseN horse_N)) (UseV play_V)
+LangEng: which horses play
+LangEst: mis hobused mängivad
+
+-- With numeral
+
+-- Number 1 doesn't behave like numerals
+Lang: QuestVP (IdetCN (IdetQuant which_IQuant (NumCard (NumNumeral (num (pot2as3 (pot1as2 (pot0as1 pot01))))))) (UseN horse_N)) (UseV play_V)
+LangEng: which one horse plays
+LangEst: mis üks hobune mängib
+
+
+Lang: QuestVP (IdetCN (IdetQuant which_IQuant (NumCard (NumNumeral (num (pot2as3 (pot1as2 (pot0as1 (pot0 n8)))))))) (UseN horse_N)) (UseV play_V)
+LangEng: which eight horses play
+LangEst: mis kaheksa hobust mängivad