Start numerical factors

src/GEval/BlackBoxDebugging.hs

@@ -7,5 +7,6 @@ data BlackBoxDebuggingOptions = BlackBoxDebuggingOptions {
   bbdoWordShapes :: Bool,
   bbdoBigrams :: Bool,
   bbdoCartesian :: Bool,
-  bbdoMinCartesianFrequency :: Maybe Integer
+  bbdoMinCartesianFrequency :: Maybe Integer,
+  bbdoConsiderNumericalFeatures :: Bool
 }

src/GEval/FeatureExtractor.hs

@@ -7,7 +7,10 @@ module GEval.FeatureExtractor
    LineWithFeatures(..),
    LineWithPeggedFactors(..),
    PeggedFactor(..),
-   Feature(..))
+   Feature(..),
+   SimpleFactor(..),
+   AtomicFactor(..),
+   FeatureNamespace(..))
   where
 
 import Data.Text
@@ -17,6 +20,7 @@ import Text.Tokenizer
 import Text.WordShape
 import GEval.BlackBoxDebugging
 import GEval.Common
+import Text.Read (readMaybe)
 
 data LineWithFeatures = LineWithFeatures Double MetricValue [Feature]
                               deriving (Eq, Ord)
@@ -37,12 +41,14 @@ data PeggedFactor = PeggedFactor FeatureNamespace SimpleFactor
 instance Show PeggedFactor where
   show (PeggedFactor namespace factor) = (show namespace) ++ ":" ++ (show factor)
 
-data SimpleFactor = SimpleAtomicFactor AtomicFactor | BigramFactor AtomicFactor AtomicFactor
+data SimpleFactor = SimpleAtomicFactor AtomicFactor | BigramFactor AtomicFactor AtomicFactor | NumericalFactor (Maybe Double) Int
                deriving (Eq, Ord)
 
 instance Show SimpleFactor where
   show (SimpleAtomicFactor factor) = show factor
   show (BigramFactor factorA factorB) = (show factorA) ++ "++" ++ (show factorB)
+  show (NumericalFactor (Just v) _) = ("=" ++ (show v))
+  show (NumericalFactor (Nothing) l) = ("=#" ++ (show l))
 
 data AtomicFactor = TextFactor Text | ShapeFactor WordShape
                      deriving (Eq, Ord)
@@ -72,12 +78,16 @@ extractAtomicFactors mTokenizer bbdo t = [Data.List.map TextFactor tokens] ++
 
 extractSimpleFactors :: (Maybe Tokenizer) -> BlackBoxDebuggingOptions -> Text -> [SimpleFactor]
 extractSimpleFactors mTokenizer bbdo t = Data.List.concat $ (Prelude.map (Prelude.map SimpleAtomicFactor) atomss) ++
-                                                             if bbdoBigrams bbdo
+                                                             (if bbdoBigrams bbdo
                                                               then Prelude.map bigramFactors atomss
-                                                             else []
+                                                              else [])
+                                                                                                                  ++
+                                                             (if bbdoConsiderNumericalFeatures bbdo
+                                                              then [numericalFactor t]
+                                                              else [])
   where atomss = extractAtomicFactors mTokenizer bbdo t
         bigramFactors atoms = Prelude.map (\(a, b) -> BigramFactor a b) $ bigrams atoms
-
+        numericalFactor t = [NumericalFactor (readMaybe $ unpack t) (Data.Text.length t)]
 extractFactors :: (Maybe Tokenizer) -> BlackBoxDebuggingOptions -> Text -> Text -> [PeggedFactor]
 extractFactors mTokenizer bbdo namespace record =
   Prelude.map (\af -> PeggedFactor (FeatureNamespace namespace) af)

src/GEval/OptionsParser.hs

@@ -193,6 +193,9 @@ blackBoxDebuggingOptionsParser = BlackBoxDebuggingOptions
                 ( long "min-cartesian-frequency"
                   <> metavar "N"
                   <> help "When combining features into Cartesian features, consider only features whose frequency exceeds the threshold given"))
+  <*> switch
+    ( long "numerical-features"
+      <> help "Consider numerical features or field lengths")
 
 singletonMaybe :: Maybe a -> Maybe [a]
 singletonMaybe (Just x) = Just [x]

test/Spec.hs

@@ -21,6 +21,8 @@ import Options.Applicative
 import Data.Text
 import Text.EditDistance
 import GEval.Annotation
+import GEval.BlackBoxDebugging
+import GEval.FeatureExtractor
 
 import Data.Map.Strict
 
@@ -33,6 +35,8 @@ import System.IO
 import System.IO.Temp
 import System.IO.Silently
 
+import Data.List (sort)
+
 import qualified Test.HUnit as HU
 
 import qualified Data.IntSet as IS
@@ -507,6 +511,36 @@ main = hspec $ do
           token <- readToken
           return $ token == (Just "BBBB")
        ) `shouldReturn` True
+  describe "extracting features" $ do
+    it "extract factors" $ do
+      let bbdo = BlackBoxDebuggingOptions {
+         bbdoMinFrequency = 1,
+         bbdoWordShapes = False,
+         bbdoBigrams = True,
+         bbdoCartesian = False,
+         bbdoMinCartesianFrequency = Nothing,
+         bbdoConsiderNumericalFeatures = True }
+      (sort $ extractFactorsFromTabbed Nothing bbdo "in" "I like this\t34.3\ttests") `shouldBe` [
+         PeggedFactor (FeatureTabbedNamespace "in" 1)
+                      (SimpleAtomicFactor (TextFactor "I")),
+         PeggedFactor (FeatureTabbedNamespace "in" 1)
+                      (SimpleAtomicFactor (TextFactor "like")),
+         PeggedFactor (FeatureTabbedNamespace "in" 1)
+                      (SimpleAtomicFactor (TextFactor "this")),
+         PeggedFactor (FeatureTabbedNamespace "in" 1)
+                      (BigramFactor (TextFactor "I") (TextFactor "like")),
+         PeggedFactor (FeatureTabbedNamespace "in" 1)
+                      (BigramFactor (TextFactor "like") (TextFactor "this")),
+         PeggedFactor (FeatureTabbedNamespace "in" 1)
+                      (NumericalFactor Nothing 11),
+         PeggedFactor (FeatureTabbedNamespace "in" 2)
+                      (SimpleAtomicFactor (TextFactor "34.3")),
+         PeggedFactor (FeatureTabbedNamespace "in" 2)
+                      (NumericalFactor (Just 34.3) 4),
+         PeggedFactor (FeatureTabbedNamespace "in" 3)
+                      (SimpleAtomicFactor (TextFactor "tests")),
+         PeggedFactor (FeatureTabbedNamespace "in" 3)
+                      (NumericalFactor Nothing 5) ]
 
 checkConduitPure conduit inList expList = do
   let outList = runConduitPure $ CC.yieldMany inList .| conduit .| CC.sinkList