Merge pull request #70 from inariksit/arabic

Arabic

src/arabic/MissingAra.gf

@@ -49,6 +49,7 @@ oper ReflA2 : A2 -> AP = notYet "ReflA2" ;
 oper ReflVP : VPSlash -> VP = notYet "ReflVP" ;
 oper SentAP : AP -> SC -> AP = notYet "SentAP" ;
 oper SentCN : CN -> SC -> CN = notYet "SentCN" ;
+oper SlashPrep : Cl -> Prep -> ClSlash = notYet "SlashPrep" ;
 oper Slash2V3 : V3 -> NP -> VPSlash = notYet "Slash2V3" ;
 oper SlashV2A : V2A -> AP -> VPSlash = notYet "SlashV2A" ;
 oper SlashV2Q : V2Q -> QS -> VPSlash = notYet "SlashV2Q" ;

src/arabic/NounAra.gf

@@ -186,7 +186,7 @@ lin
   ComplN3 n3 np = ComplN2 n3 np ** {c2 = n3.c3} ;
 
   AdjCN ap cn = cn ** {
-    adj = \\n,d,c => ap.s ! cn.h ! cn.g ! n ! (definite ! d) ! c 
+    adj = \\n,d,c => cn.adj ! n ! d ! c ++ ap.s ! cn.h ! cn.g ! n ! (definite ! d) ! c 
     };
 
   RelCN cn rs = cn ** {s = \\n,s,c => cn.s ! n ! s ! c ++ rs.s ! {pgn=Per3 cn.g n ; isPron=False} ! c};

src/arabic/ParadigmsAra.gf

@@ -374,15 +374,15 @@ resource ParadigmsAra = open
 
   v1' : Str ->  Vowel -> Vowel -> Verb =
     \rootStr,vPerf,vImpf ->
-    let { root = mkRoot3 rootStr } in 
+    let root = mkRoot3 rootStr
-    case <root.l, root.c> of {
+     in case rootStr of {
-      <"ّ",    _> => v1geminate rootStr vPerf vImpf ;
+          _          + "ّ"   => v1geminate rootStr vPerf vImpf ;
-      <"و"|"ي",_> => case vPerf of {
+          ? + #hamza + #weak => v1doubleweak root ;
+          ? + ?      + #weak => case vPerf of {
                                   i => v1defective_i root vImpf ;
                                   _ => v1defective_a root vImpf } ;
-      <_,"و"|"ي"> => v1hollow root vImpf ;
+          ? + #weak + ?      => v1hollow root vImpf ;
-      _           => v1sound root vPerf vImpf
+          _                  => v1sound root vPerf vImpf } ;
-    };
 
   v2 =
     \rootStr ->
@@ -536,6 +536,8 @@ resource ParadigmsAra = open
       lock_NP = <>
     };
 
+  proDrop : NP -> NP = ResAra.proDrop ; -- Force a NP to lose its string, only contributing with its agreement.
+
   -- e.g. al-jamii3, 2a7ad
   regNP : Str -> Number -> NP = \word,n ->
     { s = \\c => fixShd word (dec1sg ! Def ! c) ;

src/arabic/QuestionAra.gf

@@ -42,10 +42,11 @@ concrete QuestionAra of Question = CatAra ** open ResAra, ParamX, Prelude, VerbA
    CompIAdv iadv = { s = \\_ => iadv.s ; a = ResAra.Sg } ;
 
    --  QCl = {s : Tense => Polarity => QForm => Str} ;
-   QuestSlash ip cl = { ----IL just guessing
+   QuestSlash ip cls = { ----IL just guessing
       s = \\t,p,qf => 
-         let o = case qf of { QDir => Nominal ; _ => Verbal } ; -- purely guessing
+        let cl : ResAra.Cl = complClSlash cls ; -- dummy conversion to Cl
-          in cl.c2.s ++ ip.s ! False ! Def ! Nom ++ cl.s ! t ! p ! o
+            o = case qf of { QDir => Nominal ; _ => Verbal } ; -- purely guessing
+          in cls.c2.s ++ ip.s ! False ! Def ! Nom ++ cl.s ! t ! p ! o
       } ;
 
   --IL guessed

src/arabic/RelativeAra.gf

@@ -20,12 +20,13 @@ concrete RelativeAra of Relative = CatAra **
      } ;
 
     -- : RP -> ClSlash -> RCl ; -- whom John loves 
-   RelSlash rp cl = cl ** {
+   RelSlash rp cls = cls ** {
      s = \\t,p,agr,c => 
         let obj = case (pgn2gn agr.pgn).g of {
-        			Fem  => St.she_Pron ;
+        			Fem  => St.she_Pron ; -- head is repeated as a clitic object pronoun
         			Masc => St.he_Pron } ;
-        in rp.s ! agr2ragr agr c ++ cl.s ! t ! p ! Nominal ++ cl.c2.s ++ obj.s ! cl.c2.c
+            cl : ResAra.Cl = complClSlash obj cls ;
+        in rp.s ! agr2ragr agr c ++ cl.s ! t ! p ! VOS
      } ;
 --
 --    FunRP p np rp = {

src/arabic/ResAra.gf

@@ -26,7 +26,8 @@ resource ResAra = PatternsAra ** open  Prelude, Predef, OrthoAra, ParamX  in {
                      -- case vowel retained
     Mood    = Ind | Cnj | Jus ;
     Voice   = Act | Pas ;
-    Order   = Verbal | Nominal ;
+    Order   = Verbal | Nominal
+            | VOS ; -- Relative clauses with resumptive pronouns
 
   oper
 
@@ -552,6 +553,12 @@ v1defective_i : Root3 -> Vowel -> Verb = \bqy,vowImpf -> -- IL (conjugation 1d4)
                          x => vforms_a ! x } ;
    in verbDef vforms_i vowImpf ;
 
+v1doubleweak : Root3 -> Verb = \r'y ->
+  let ry = r'y ** {c = ""} ;
+      vforms : DefForms = \\x => rmSukun (v1DefForms_perfA ry a ! x) ; -- only remove the first sukun
+   in verbDoubleDef vforms i ; -- sukun in suffixes is removed in verbDoubleDef
+
+
 patDef1 : Vowel => Pattern =
   table {
     u => fcu ;
@@ -641,7 +648,6 @@ v4doubleweak : Root3 -> Verb = \r'y ->
       vforms : DefForms = \\x => rmSukun (v4DefForms ry ! x) ; -- only remove the first sukun
    in verbDoubleDef vforms i ; -- sukun in suffixes is removed in verbDoubleDef
 
-
 v5sound : Root3 -> Verb =
   \nfs ->
   let {
@@ -1181,6 +1187,8 @@ patHollowImp : (_,_ :Str) -> Gender => Number => Str =\xaf,xAf ->
     Agr = { pgn : PerGenNum; isPron : Bool} ;
     AAgr = { g : Gender ; n : Number} ;
 
+
+
     Comp : Type = {
       s : AAgr => Case => Str
       } ;
@@ -1201,6 +1209,14 @@ patHollowImp : (_,_ :Str) -> Gender => Number => Str =\xaf,xAf ->
       a : Agr
       } ;
 
+    proDrop : NP -> NP = \np -> 
+      case np.a.isPron of {
+        True => np ** {s = \\_ => []};
+        _    => np
+      } ;
+
+    emptyNP : NP = {s = \\_ => [] ; a = {pgn = Per3 Masc Sg ; isPron = False}} ;
+
     IP : Type = {
       s : Bool -- different forms for "what is this" and "what do you do"
        => State => Case   -- because of PrepIP: e.g. "in which" chooses definite accusative
@@ -1274,8 +1290,48 @@ patHollowImp : (_,_ :Str) -> Gender => Number => Str =\xaf,xAf ->
       };
 
 
-   predVSlash : Verb2 -> VPSlash = \v ->
+    predVP : NP -> VP -> Cl = \np,vp ->
-     predV v ** {c2 = v.c2} ;
+      { s =\\t,p,o =>
+          let {
+            pgn =
+              case <o,np.a.isPron> of {
+                <Verbal, False> => verbalAgr np.a.pgn;
+                _               => np.a.pgn
+              };
+            gn = pgn2gn pgn;
+            kataba  = vp.s ! pgn ! VPPerf ;
+            yaktubu = vp.s ! pgn ! VPImpf Ind ;
+            yaktuba = vp.s ! pgn ! VPImpf Cnj ;
+            yaktub  = vp.s ! pgn ! VPImpf Jus ;
+            vStr : ParamX.Tense -> Polarity -> Str =
+              \tn,pl -> case <vp.isPred,tn,pl> of {
+              <False, Pres, Pos> => yaktubu ;
+              <False, Pres, Neg> => "لَا" ++ yaktubu ;
+              <True, Pres, Pos> => "" ;      --no verb "to be" in present
+              <True, Pres, Neg> => "لَيسَ" ;--same here, just add negation particle
+              <_, Past, Pos> => kataba ;
+              <_, Past, Neg> => "لَمْ" ++ yaktub ;
+              <_, _Fut,  Pos> => "سَ" ++ yaktubu ;
+              <_, _Fut,  Neg> => "لَنْ" ++ yaktuba
+              };
+            pred : ParamX.Tense -> Polarity -> Str =
+              \tn,pl -> case <vp.isPred,tn,pl>  of {
+              <True, Pres, Pos> => vp.pred.s ! gn ! Nom; --xabar marfooc
+              _ => vp.pred.s ! gn ! Acc --xabar kaana wa laysa manSoob
+              };
+
+          } in
+          -- If you want prodrop, use proDrop : NP -> NP for your subject. /IL
+          case o of {
+            Verbal => vStr t p ++ case vp.obj.a.isPron of {
+                            True  => vp.obj.s ++ np.s ! Nom ; -- obj. clitic attaches directly to the verb                                   
+                            False => np.s ! Nom ++ vp.obj.s }
+                   ++ vp.s2 ++ pred t p ;
+            Nominal => np.s ! Nom ++ vStr t p ++ vp.obj.s ++ vp.s2 ++ pred t p ;
+            VOS => vStr t p ++ vp.obj.s ++ vp.s2 ++ pred t p ++ np.s ! Nom
+
+          }
+      } ;
 
     -- in verbal sentences, the verb agrees with the subject
     -- in Gender but not in number
@@ -1297,11 +1353,30 @@ patHollowImp : (_,_ :Str) -> Gender => Number => Str =\xaf,xAf ->
       { s2 = vp.s2 ++ str };
 
     kaan : {s : AAgr => Case => Str} -> VP = \xabar ->
-      insertPred xabar (predV (v1hollow {f = "ك"; c = "و" ; l = "ن"} u) );
+      insertPred xabar (predV copula);
 
+    copula : Verb = v1hollow {f = "ك"; c = "و" ; l = "ن"} u ;
     -- Slash categories
+
     VPSlash : Type = VP ** {c2 : Preposition} ;
-    ClSlash : Type = Cl ** {c2 : Preposition} ;
+    ClSlash : Type = VPSlash ** {subj : NP} ;
+
+    slashV2 : Verb2 -> VPSlash = \v ->
+      predV v ** {c2 = v.c2} ;
+
+    -- Add subject string, fix agreement to the subject,
+    -- but keep the structure as VP, because later on
+    -- we might need different word orders for the ClSlash.
+    predVPSlash : NP -> VPSlash -> ClSlash = \np,v -> v ** {
+      subj = np
+      } ;
+    
+    complClSlash = overload {
+      complClSlash : NP -> ClSlash -> Cl = \obj,cls ->
+        predVP cls.subj (insertObj obj cls) ;
+      complClSlash :       ClSlash -> Cl = \cls ->  
+        predVP cls.subj (insertObj emptyNP cls) -- Empty subject and object
+      } ;
 
     Cl  : Type = {s : Tense => Polarity => Order => Str} ;
     QCl : Type = {s : Tense => Polarity => QForm => Str} ;
@@ -1315,9 +1390,8 @@ patHollowImp : (_,_ :Str) -> Gender => Number => Str =\xaf,xAf ->
       agr2ragr : Agr -> Case -> RAgr = \a,c ->
         let gn = pgn2gn a.pgn in case <gn.n,gn.g,a> of {
           <Sg,x> => RSg x ;
-          <Pl,x> => RPl x ;
           <Dl,x> => RDl x c ;
-          _      => Predef.error "agr2ragr"} ;
+          <Pl,x> => RPl x } ;
       agr2ragr : Number -> Case -> Gender -> RAgr = \n,c,g ->
         case n of {
           Sg => RSg g ;

src/arabic/SentenceAra.gf

@@ -26,58 +26,7 @@ concrete SentenceAra of Sentence = CatAra ** open
           }
       };
 -}
-    PredVP np vp =
+    PredVP = predVP ;
-      { s =\\t,p,o =>
-          let {
-            pgn =
-              case <o,np.a.isPron> of {
-                <Verbal, False> => verbalAgr np.a.pgn;
-                _               => np.a.pgn
-              };
-            gn = pgn2gn pgn;
-            kataba  = vp.s ! pgn ! VPPerf ;
-            yaktubu = vp.s ! pgn ! VPImpf Ind ;
-            yaktuba = vp.s ! pgn ! VPImpf Cnj ;
-            yaktub  = vp.s ! pgn ! VPImpf Jus ;
-            vStr : ParamX.Tense -> Polarity -> Str =
-              \tn,pl -> case<vp.isPred,tn,pl> of {
-              <False, Pres, Pos> => yaktubu ;
-              <False, Pres, Neg> => "لَا" ++ yaktubu ;
-              <True, Pres, Pos> => "" ;      --no verb "to be" in present
-              <True, Pres, Neg> => "لَيسَ" ;--same here, just add negation particle
-              <_, Past, Pos> => kataba ;
-              <_, Past, Neg> => "لَمْ" ++ yaktub ;
-              <_, _Fut,  Pos> => "سَ" ++ yaktubu ;
-              <_, _Fut,  Neg> => "لَنْ" ++ yaktuba
-              };
-            pred : ParamX.Tense -> Polarity -> Str =
-              \tn,pl -> case <vp.isPred,tn,pl>  of {
-              <True, Pres, Pos> => vp.pred.s ! gn ! Nom; --xabar marfooc
-              _ => vp.pred.s ! gn ! Acc --xabar kaana wa laysa manSoob
-              };
-
-          } in
-          case o of {
-            Verbal =>
-              --case <False, np.a.isPron> of { ---- AR workaround 18/12/2008
-          case <vp.obj.a.isPron, np.a.isPron> of {
-        {- IL: I don't think we should do prodrop here. vStr drops the copula in present tense,
-               so there's hardly anything left for a predicative clause: e.g.
-                  PredVP (UsePron i_Pron) (UseComp (CompCN (UseN car_N))) "I am a car"
-               would be linearised just as "car", if we have both prodrop and copula drop.
-               Leaving it up to someone who knows Arabic to decide what is better.
-        Original here:
-                <True,True>  => (vStr t p) ++ vp.obj.s ++ vp.s2 ++ (pred t p) ;
-                -- ya2kuluhu
-                <False,True> => (vStr t p) ++ vp.obj.s ++ vp.s2 ++ (pred t p); -}
-                -- ya2kuluhu al-waladu, yakuluhu al-2awlaadu
-                <False> => (vStr t p) ++ np.s ! Nom ++ vp.obj.s ++ vp.s2 ++ (pred t p);
-                <True>  => (vStr t p) ++ vp.obj.s ++ np.s ! Nom ++ vp.s2 ++ (pred t p)
-              };
-            Nominal =>
-              np.s ! Nom ++ (vStr t p) ++ vp.obj.s ++ vp.s2 ++ (pred t p)
-          }
-      };
 
 --    PredSCVP sc vp = mkClause sc.s (agrP3 Sg) vp ;
 
@@ -100,9 +49,13 @@ concrete SentenceAra of Sentence = CatAra ** open
 
 -- ClSlash
 
-    SlashVP np vps = PredVP np vps ** { c2 = vps.c2 } ;
+    SlashVP = predVPSlash ;
     AdvSlash slash adv = slash ** { s2 = slash.s2 ++ adv.s } ;
-    SlashPrep cl prep = cl ** {c2 = prep} ;
+
+-- SlashPrep : Cl -> Prep -> ClSlash
+-- Will be awkward to implement in the way ClSlash is now.
+-- ClSlash is implemented the way it is now for a good reason:
+-- we need to support different word orders.
 
 --  SlashVS np vs sslash = TODO
 
@@ -111,8 +64,8 @@ concrete SentenceAra of Sentence = CatAra ** open
 --    EmbedQS qs = {s = qs.s ! QIndir} ;
 --    EmbedVP vp = {s = infVP False vp (agrP3 Sg)} ; --- agr
 --
-    UseSlash,
+
-    UseCl = \t,p,cl ->
+    UseCl t p cl =
       {s = t.s ++ p.s ++
          case <t.t,t.a> of { --- IL guessed tenses
              <(Pres|Cond),Simul>  => cl.s ! Pres ! p.p ! Nominal ;
@@ -132,4 +85,5 @@ concrete SentenceAra of Sentence = CatAra ** open
 
     UseRCl t p cl = {s = \\agr,c => t.s ++ p.s ++ cl.s ! t.t ! p.p ! agr ! c} ;
 
+    UseSlash t p cl = UseCl t p (complClSlash cl) ;
 }

src/arabic/VerbAra.gf

@@ -11,8 +11,8 @@ concrete VerbAra of Verb = CatAra ** open Prelude, ResAra in {
     --SlashV2V : V2V -> VP -> VPSlash ;  -- beg (her) to go
     --SlashV2VNP : V2V -> NP -> VPSlash -> VPSlash ; -- beg me to buy
 
-    SlashV2a v = predVSlash v ;
+    SlashV2a = slashV2 ;
-    Slash3V3 v np = insertObj np (predVSlash v) ** {c2 = v.c3};
+    Slash3V3 v np = insertObj np (slashV2 v) ** {c2 = v.c3};
 
     ComplSlash vp np = insertObj np vp ;
 
@@ -38,6 +38,8 @@ concrete VerbAra of Verb = CatAra ** open Prelude, ResAra in {
 --
     UseComp xabar = kaan xabar ;
 
+    UseCopula = predV copula ;
+
     AdvVP vp adv = insertStr adv.s vp ;
 
 --    AdVVP adv vp = insertAdV adv.s vp ;