diff --git a/lib/src/latin/dict/MkLatin.hs b/lib/src/latin/dict/MkLatin.hs
new file mode 100644
index 000000000..480c54772
--- /dev/null
+++ b/lib/src/latin/dict/MkLatin.hs
@@ -0,0 +1,60 @@
+import Data.Char
+
+main = mkLatin
+
+mkLatin = do
+  ws <- readFile "DICTPAGE.RAW" >>= return . lines
+  let fcs = mkDict ws
+  let fs = [unwords [status st ++ "fun",f,":",cat,";","--",eng] | [st, f,cat,_,eng] <- fcs]
+  let ls = [unwords [status st ++ "lin",f,"=",lat,";"]          | [st, f,_,lat,_] <- fcs]
+  
+  writeFile  "DictLatAbs.gf" "abstract DictLatAbs = Cat ** {"
+  appendFile "DictLatAbs.gf" "\n-- extracted from http://archives.nd.edu/whitaker/dictpage.htm\n"
+  appendFile "DictLatAbs.gf" (unlines fs)
+  appendFile "DictLatAbs.gf" "}"
+  
+  writeFile  "DictLat.gf" "concrete DictLat of DictLatAbs = CatLat ** open ParadigmsLat in {"
+  appendFile "DictLat.gf" "\n-- extracted from http://archives.nd.edu/whitaker/dictpage.htm\n"
+  appendFile "DictLat.gf" (unlines ls)
+  appendFile "DictLat.gf" "}"
+
+---  putStrLn $ unlines ls
+
+mkDict :: [String] -> [[String]] -- fun, cat, lat, eng
+mkDict = map mkOne . zip [10001 ..] . map cleanUp 
+  where
+   cleanUp s = let (lat,eng) = break (=='[') s in 
+             (words (filter (\c -> c==' ' || isLetter c) lat), eng)
+   mkOne (i,(lws,eng)) = addId i (mkLat lws) ++ [eng]
+   mkLat lws = case lws of 
+     x:y:"N":_:g:_ -> f [x, "N", lin "mkN" [show x,show y, (snd (gender g))]] where f = if fst (gender g) then ok else todo
+     x:y:"N":  g:_ -> f [x, "N", lin "mkN" [show x,show y, (snd (gender g))]] where f = if fst (gender g) then ok else todo
+     x:"gen":z:"ADJ":_ -> ok [x, "A", lin "mkA" [show x]]
+     x:y:z:"ADJ":_ -> ok [x, "A", lin "mkA" [show x]]
+     x:"ADV":_     -> ok [x, "Adv", lin "mkAdv" [show x]]
+     x:y:z:u:"V":_:"INTRANS":_ -> okv   [y, "V",                     lin "mkV" [show y,show x,show z,show u]]
+     x:y:z:u:"V":_:"TRANS":_   -> okv   [y, "V2", lin "mkV2" ["(" ++ lin "mkV" [show y,show x,show z,show u] ++ ")"]]
+     x:y:z:u:"V":_:"DEP":_     -> todo [y, "V",  lin "depV" ["(" ++ lin "mkV" [show y,show x,show z,show u] ++ ")"]]
+     x:y:z:u:"V":_             -> okv   [y, "V",                     lin "mkV" [show y,show x,show z,show u]]
+
+     _ -> todo ["TODO","",unwords lws]
+
+   addId i (st:f:c:rest) = st:(f ++ "_" ++ show i ++ "_" ++ c):c:rest
+   
+   gender g = case g of 
+     "M" -> (True, "masculine")
+     "F" -> (True, "feminine")
+     "N" -> (True, "neuter")
+     "C" -> (True, "masculine {-C-}")
+     _   -> (False, g ++ "{-??-}")
+
+   fun x c = x ++ "_" ++ c
+   lin f xs = unwords (f:xs)
+   todo xs = "1":xs
+   ok xs = "0":xs
+   okv r@(f:_) = if elem (take 3 (reverse f)) ["era","ere","eri"] then ok r else todo r
+
+status st = case st of
+  "0" -> ""
+  _ -> "-- "
+
diff --git a/lib/src/latin/dict/README b/lib/src/latin/dict/README
new file mode 100644
index 000000000..372a21ee0
--- /dev/null
+++ b/lib/src/latin/dict/README
@@ -0,0 +1,33 @@
+Aarne Ranta 2017-01-19
+
+A script for building a Latin dictionary from William Whitaker's Words
+
+Source: http://archives.nd.edu/whitaker/dictpage.zip
+
+See also: https://en.wikipedia.org/wiki/William_Whitaker's_Words
+
+License: BSD for the script. Unknown for the source, therefore not redistributed here.
+
+Usage:
+
+  runghc MkLatin.hs
+
+Output:
+
+  DictLatAbs.gf
+  DictLat.gf
+
+Uses Herbert Lange's RGL implementation of Latin (darcs version of 2017-01-19 with ParadigmsLat.mkAdv added by AR)
+
+Coverage of first version: 90%; 3938 missing out of 39227 entries. Marked with leading "--"
+Reasons for uncovered entries
+- uncommon category (e.g. structural words)
+- missing case in ParadigmsLat.gf, e.g. deponent verbs
+
+Matching for adjectives (A) uncertain, because we only use ParadigmsLat.gf.mkA : Str -> A
+
+Verb subcat information is uncertain, too. It is based on the src's TRANS and INTRANS flags.
+
+
+
+