Started experimenting with shallow resource API.

Started experimenting with shallow resource API.
Added PossessPrep to Structural.

lib/resource-0.6/shallow/lexicon/MkLex.hs

@@ -0,0 +1,66 @@
+module MkLex where
+
+import System
+import Monad
+
+-- to massage an ispell word list into a GF lexicon. AR 20/2/2004
+
+lexfile = "linux.words"
+absfile = "Lex.gf"
+cncfile = "LexEng.gf"
+maxsize = 10000 :: Int --- add more!
+
+massage :: IO ()
+massage = do
+  -- initialize target files
+  system $ "echo \"\" >" ++ absfile
+  system $ "echo \"\" >" ++ cncfile
+  appendFile absfile $ 
+    "abstract Lex = Shallow ** {\nfun\n"
+  appendFile cncfile $ 
+    "--# -path=.:..:../../../prelude:../../abstract:../../english\n\n"
+  appendFile cncfile $ 
+    "concrete LexEng of Lex = ShallowEng ** open ParadigmsEng in {\nlin\n"
+  -- reverse to study endings
+  ws <- liftM (map reverse . lines) $ readFile lexfile
+  sortWords $ take maxsize ws
+
+-- we exploit the fact that the original list is sorted and
+-- different forms therefore lie consecutively
+sortWords :: [String] -> IO ()
+sortWords ws = case ws of
+  u : ('d':'e':v) : ('s':w) : vs | v == u && w == u -> -- regular verb
+    mkEntry verbReg u >> sortWords vs
+  ('e':u) : ('d':'e':v) : ('s':'e':w) : vs | v == u && w == u -> -- e-verb
+    mkEntry verbE ('e':u) >> sortWords vs
+  u : ('y':'l':v) : vs | v == u ->                     -- regular adjective
+    mkEntry adjReg u >> sortWords vs
+  u : ('s':v) : vs | v == u ->                         -- regular noun
+    mkEntry nounReg u >> sortWords vs
+  ---- add more
+  _ : vs -> sortWords vs
+  [] -> appendFile absfile "  }\n" >> appendFile cncfile "  }\n"
+
+mkEntry :: (String -> (String,String)) -> String -> IO ()
+mkEntry abc w0 = do
+  let w = reverse w0
+  let (ab,cn) = abc w
+  appendFile absfile $ "  " ++ ab
+  appendFile absfile $ " ;\n"
+  appendFile cncfile $ "  " ++ cn
+  appendFile cncfile $ " ;\n"
+
+verbReg :: String -> (String,String)
+verbReg = mkGF "Verb" "vReg"
+
+verbE = verbReg ----
+adjReg  = mkGF "Adj" "mkAdj1"
+nounReg = mkGF "Noun" "cnNonhuman"
+
+
+mkGF :: String -> String -> (String -> (String,String))
+mkGF cat oper w = (
+  w ++ "L : " ++ cat,
+  w ++ "L = " ++ oper ++ " \"" ++ w ++ "\""
+  )
+

lib/resource-0.6/shallow/lexicon/notes.txt

@@ -0,0 +1,31 @@
+Project: to massage an ispell word list into a GF lexicon. AR 20/2/2004
+
+Used MkLex on linux.words (45424 word forms)
+
+The first run gave 6471 words, which was too much for GF.
+With the maxsize, got this down to 1587.
+The grammar now compiles in 25sec, and parses as follows:
+
+  "every commander assumes of an abbot"                    
+   610 msec, 6 ws
+
+  "every accusing commander assumes of a coarse abbot"     
+   840 msec, 8 ws
+
+  "every accusing commander assumes of a coarse abbot of all 
+   cognitive colonizers"
+   820 msec, 12 ws
+
+  "every accusing commander of a comical careful colt assumes of 
+   a coarse abbot of all cognitive colonizers"
+   3670 msec, 17 ws (nesting inside)
+
+  "every accusing commander assumes of a coarse abbot of all cognitive 
+   colonizers of a colt of a careful carpenter of every carriage"
+   TIMEOUT!
+
+Note: it is essential to import with the -cflexer flag!
+
+Time spent: 58 min, includes writing MkLex, running tests, 
+and writing these notes.
+