concraft 0.6.0 → 0.7.0

raw patch · 3 files changed

+119/−2 lines, 3 filesPVP ok

version bump matches the API change (PVP)

API changes (from Hackage documentation)

Files

• concraft.cabal +3/−2
• src/NLP/Concraft/Morphosyntax/Accuracy.hs +109/−0
• src/NLP/Concraft/Morphosyntax/Align.hs +7/−0

@@ -1,5 +1,5 @@ name:               concraft-version:            0.6.0+version:            0.7.0 synopsis:           Morphological disambiguation based on constrained CRFs description:     A morphological disambiguation library based on@@ -47,11 +47,12 @@      exposed-modules:         NLP.Concraft-      , NLP.Concraft.Morphosyntax       , NLP.Concraft.Analysis       , NLP.Concraft.Schema       , NLP.Concraft.Guess       , NLP.Concraft.Disamb+      , NLP.Concraft.Morphosyntax+      , NLP.Concraft.Morphosyntax.Accuracy      other-modules:         NLP.Concraft.Disamb.Positional

@@ -0,0 +1,109 @@+-- | Accuracy statistics.+++module NLP.Concraft.Morphosyntax.Accuracy+(+-- * Stats+  Stats (..)+, accuracy++-- * Accuracy+, weakLB+, weakUB+, strongLB+, strongUB+) where +++import           Data.List (foldl')+import qualified Data.Set as S+import qualified Data.Map as M+import qualified Data.Tagset.Positional as P+import           NLP.Concraft.Morphosyntax+import           NLP.Concraft.Morphosyntax.Align+++-- | Statistics.+data Stats = Stats+    { good :: Int   -- ^ Number of correct tags+    , gold :: Int } -- ^ Number of segments in gold corpus+++-- | Add stats,+(.+.) :: Stats -> Stats -> Stats+Stats x y .+. Stats x' y' = Stats (x + x') (y + y')+++-- | Accuracy given stats.+accuracy :: Stats -> Double+accuracy s+    = fromIntegral (good s)+    / fromIntegral (gold s)+++-- | Accuracy weak lower bound.+weakLB :: Word w => P.Tagset -> [Seg w P.Tag] -> [Seg w P.Tag] -> Stats+weakLB tagset ref other =+    foldl' (.+.) (Stats 0 0) . map (uncurry stats) $ align ref other+  where+    stats [x] [y]+        | S.null (xTags `S.intersection` yTags) = Stats 0 1+        | otherwise = Stats 1 1+      where+        xTags = choice tagset x+        yTags = choice tagset y+    stats xs _ = Stats 0 (length xs)+++-- | Accuracy strong lower bound.+strongLB :: Word w => P.Tagset -> [Seg w P.Tag] -> [Seg w P.Tag] -> Stats+strongLB tagset ref other =+    foldl' (.+.) (Stats 0 0) . map (uncurry stats) $ align ref other+  where+    stats [x] [y]+        | xTags == yTags = Stats 1 1+        | otherwise = Stats 0 1+      where+        xTags = choice tagset x+        yTags = choice tagset y+    stats xs _ = Stats 0 (length xs)+++-- | Accuracy weak upper bound.+weakUB :: Word w => P.Tagset -> [Seg w P.Tag] -> [Seg w P.Tag] -> Stats+weakUB tagset ref other =+    foldl' (.+.) (Stats 0 0) . map (uncurry stats) $ align ref other+  where+    stats [x] [y]+        | S.null (xTags `S.intersection` yTags) = Stats 0 1+        | otherwise = Stats 1 1+      where+        xTags = choice tagset x+        yTags = choice tagset y+    stats xs _ = Stats (length xs) (length xs)+++-- | Accuracy strong upper bound.+strongUB :: Word w => P.Tagset -> [Seg w P.Tag] -> [Seg w P.Tag] -> Stats+strongUB tagset ref other =+    foldl' (.+.) (Stats 0 0) . map (uncurry stats) $ align ref other+  where+    stats [x] [y]+        | xTags == yTags = Stats 1 1+        | otherwise = Stats 0 1+      where+        xTags = choice tagset x+        yTags = choice tagset y+    stats xs _ = Stats (length xs) (length xs)+    ++-- | All tags are expanded here.+choice :: P.Tagset -> Seg w P.Tag -> S.Set P.Tag+choice tagset = S.fromList . concatMap (P.expand tagset) . positive+++-- | Positive tags.+positive :: Seg w t -> [t]+positive seg =+    let xs = M.toList . unWMap . tags+    in  [x | (x, v) <- xs seg, v > 0]

@@ -1,12 +1,15 @@ {-# LANGUAGE TupleSections #-} + -- | Alignment and synchronization.  Currently works only with positional tagsets. + module NLP.Concraft.Morphosyntax.Align ( align , sync ) where + import           Control.Applicative ((<|>)) import           Data.Maybe (fromJust) import           Data.List (find)@@ -18,6 +21,7 @@  import           NLP.Concraft.Morphosyntax + -- | Synchronize two datasets, taking disamb tags from the first one -- and the rest of information form the second one. -- In case of differences in token-level segmentation, reference segmentation@@ -26,6 +30,7 @@ sync :: Word w => P.Tagset -> [Seg w P.Tag] -> [Seg w P.Tag] -> [Seg w P.Tag] sync tagset xs ys = concatMap (uncurry (moveDisamb tagset)) (align xs ys) + -- | If both arguments contain only one segment, insert disamb interpretations -- from the first segment into the second segment.  Otherwise, the first list -- of segments will be returned unchanged.@@ -50,6 +55,7 @@ -- Do nothing in this case. moveDisamb _ xs _ = xs + -- | Align two lists of segments. align :: Word w => [Seg w t] -> [Seg w t] -> [([Seg w t], [Seg w t])] align [] [] = []@@ -59,6 +65,7 @@     let (x, y) = match xs ys         rest   = align (drop (length x) xs) (drop (length y) ys)     in  (x, y) : rest+  -- | Find the shortest, length-matching prefixes in the two input lists. match :: Word w => [Seg w t] -> [Seg w t] -> ([Seg w t], [Seg w t])