src/Gargantext/Text/Metrics/Occurrences.hs

   1 {-|
   2 Module      : Gargantext.Text.Metrics.Occurrences
   3 Description :
   4 Copyright   : (c) CNRS, 2017-Present
   5 License     : AGPL + CECILL v3
   6 Maintainer  : team@gargantext.org
   7 Stability   : experimental
   8 Portability : POSIX
   9
  10 Token and occurrence
  11
  12 An occurrence is not necessarily a token. Considering the sentence:
  13 "A rose is a rose is a rose". We may equally correctly state that there
  14 are eight or three words in the sentence. There are, in fact, three word
  15 types in the sentence: "rose", "is" and "a". There are eight word tokens
  16 in a token copy of the line. The line itself is a type. There are not
  17 eight word types in the line. It contains (as stated) only the three
  18 word types, 'a', 'is' and 'rose', each of which is unique. So what do we
  19 call what there are eight of? They are occurrences of words. There are
  20 three occurrences of the word type 'a', two of 'is' and three of 'rose'.
  21 Source : https://en.wikipedia.org/wiki/Type%E2%80%93token_distinction#Occurrences
  22
  23 -}
  24
  25 {-# LANGUAGE NoImplicitPrelude #-}
  26 {-# LANGUAGE OverloadedStrings #-}
  27
  28 module Gargantext.Text.Metrics.Occurrences
  29   where
  30
  31
  32 import Control.Arrow ((***))
  33 import qualified Data.List as List
  34 import Data.Map.Strict  (Map
  35                         , empty, singleton
  36                         , insertWith, insertWithKey, unionWith
  37                         , toList, lookup, mapKeys
  38                         )
  39 import Data.Set (Set)
  40 import qualified Data.Set as Set
  41 import Data.Text (pack)
  42
  43 import qualified Data.Map.Strict as DMS
  44 import Control.Monad ((>>),(>>=))
  45 import Data.String (String())
  46 import Data.Attoparsec.Text
  47
  48 ------------------------------------------------------------------------
  49 import Gargantext.Prelude
  50 import Gargantext.Core.Types
  51 ------------------------------------------------------------------------
  52 type Occ  a = Map      a  Int
  53 type Cooc a = Map (a,  a) Int
  54 type FIS  a = Map (Set a) Int
  55
  56 data Group = ByStem | ByOntology
  57
  58 type Grouped = Stems
  59
  60
  61 {-
  62 -- >> let testData = ["blue lagoon", "blues lagoon", "red lagoon"]
  63 -- >> map occurrences <$> Prelude.mapM (terms Mono EN)
  64 -- [fromList [(fromList ["blue"],1),(fromList ["lagoon"],1)],fromList [(fromList ["blue"],1),(fromList ["lagoon"],1)],fromList [(fromList ["lagoon"],1),(fromList ["red"],1)]]
  65 --λ:   cooc <$> Prelude.map occurrences <$> Prelude.mapM (terms Mono EN) ["blue lagoon", "blues lagoon", "red lagoon"]
  66 --fromList [((fromList ["blue"],fromList ["lagoon"]),2),((fromList ["lagoon"],fromList ["red"]),1)]
  67 --λ:   cooc <$> Prelude.map occurrences <$> Prelude.mapM (terms Mono EN) ["blue lagoon", "blues lagoon", "red lagoon", "red lagoon"]
  68 --fromList [((fromList ["blue"],fromList ["lagoon"]),2),((fromList ["lagoon"],fromList ["red"]),2)]
  69 --λ:   cooc <$> Prelude.map occurrences <$> Prelude.mapM (terms Mono EN) ["blue lagoon", "blues lagoon", "red lagoon red lagoon", "red lagoon"]
  70 --fromList [((fromList ["blue"],fromList ["lagoon"]),2),((fromList ["lagoon"],fromList ["red"]),2)]
  71 --λ:   cooc <$> Prelude.map occurrences <$> Prelude.mapM (terms Mono EN) ["blue lagoon", "blues lagoon blues lagoon", "red lagoon red lagoon", "red lagoon"]
  72 --fromList [((fromList ["blue"],fromList ["lagoon"]),2),((fromList ["lagoon"],fromList ["red"]),2)]
  73 ----
  74
  75            -}
  76
  77 type Occs = Int
  78 type Coocs = Int
  79
  80 removeApax :: Map (Label, Label) Int -> Map (Label, Label) Int
  81 removeApax = DMS.filter (> 1)
  82
  83 cooc :: [[Terms]] -> Map (Label, Label) Int
  84 cooc tss =
  85   mapKeys (delta $ labelPolicy terms_occs) $ cooc' (map (Set.fromList . map _terms_stem) tss)
  86   where
  87     terms_occs = occurrences (List.concat tss)
  88     delta f = f *** f
  89
  90
  91 labelPolicy :: Map Grouped (Map Terms Occs) -> Grouped -> Label
  92 labelPolicy m g =  case _terms_label <$> fst <$> maximumWith snd <$> DMS.toList <$> lookup g m of
  93                      Just label -> label
  94                      Nothing    -> panic $ "Label of Grouped not found: " <> (pack $ show g)
  95
  96 cooc' :: Ord b => [Set b] -> Map (b, b) Coocs
  97 cooc' tss = foldl' (\m (xy,c) -> insertWith ((+)) xy c m) empty xs
  98   where
  99       xs = [ ((x, y), 1)
 100            | xs <- tss
 101            , ys <- tss
 102            , x <- Set.toList xs
 103            , y <- Set.toList ys
 104            , x < y
 105            ]
 106
 107
 108 -- | Compute the grouped occurrences (occ)
 109 occurrences :: [Terms] -> Map Grouped (Map Terms Int)
 110 occurrences = occurrences' _terms_stem
 111
 112 occurrences' :: (Ord a, Ord b) => (a -> b) -> [a] -> Map b (Map a Int)
 113 occurrences' f = foldl' (\m a -> insertWith (unionWith (+)) (f a) (singleton a 1) m) empty
 114
 115 -- TODO add groups and filter stops
 116 sumOcc :: Ord a => [Occ a] -> Occ a
 117 sumOcc xs = foldl' (unionWith (+)) empty xs
 118
 119