]> Git — Sourcephile - gargantext.git/blob - src/Gargantext/Text/Metrics/Count.hs
[FEAT] adding RAKE algorithm to extract multi-terms (keywords) in context of texts.
[gargantext.git] / src / Gargantext / Text / Metrics / Count.hs
1 {-|
2 Module : Gargantext.Text.Metrics.Count
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.Count
29 where
30
31 import Data.Text (Text)
32 import Control.Arrow (Arrow(..), (***))
33 import qualified Data.List as List
34
35 import qualified Data.Map.Strict as DMS
36 import Data.Map.Strict ( Map, empty, singleton
37 , insertWith, unionWith, unionsWith
38 , mapKeys
39 )
40 import Data.Set (Set)
41 import Data.Text (pack)
42
43
44 ------------------------------------------------------------------------
45 import Gargantext.Prelude
46 import Gargantext.Core.Types
47 ------------------------------------------------------------------------
48 type Occ a = Map a Int
49 type Cooc a = Map (a, a) Int
50 type FIS a = Map (Set a) Int
51
52 data Group = ByStem | ByOntology
53
54 type Grouped = Stems
55
56
57 {-
58 -- >> let testData = ["blue lagoon", "blues lagoon", "red lagoon"]
59 -- >> map occurrences <$> Prelude.mapM (terms Mono EN)
60 -- [fromList [(fromList ["blue"],1),(fromList ["lagoon"],1)],fromList [(fromList ["blue"],1),(fromList ["lagoon"],1)],fromList [(fromList ["lagoon"],1),(fromList ["red"],1)]]
61 --λ: cooc <$> Prelude.map occurrences <$> Prelude.mapM (terms Mono EN) ["blue lagoon", "blues lagoon", "red lagoon"]
62 --fromList [((fromList ["blue"],fromList ["lagoon"]),2),((fromList ["lagoon"],fromList ["red"]),1)]
63 --λ: cooc <$> Prelude.map occurrences <$> Prelude.mapM (terms Mono EN) ["blue lagoon", "blues lagoon", "red lagoon", "red lagoon"]
64 --fromList [((fromList ["blue"],fromList ["lagoon"]),2),((fromList ["lagoon"],fromList ["red"]),2)]
65 --λ: cooc <$> Prelude.map occurrences <$> Prelude.mapM (terms Mono EN) ["blue lagoon", "blues lagoon", "red lagoon red lagoon", "red lagoon"]
66 --fromList [((fromList ["blue"],fromList ["lagoon"]),2),((fromList ["lagoon"],fromList ["red"]),2)]
67 --λ: cooc <$> Prelude.map occurrences <$> Prelude.mapM (terms Mono EN) ["blue lagoon", "blues lagoon blues lagoon", "red lagoon red lagoon", "red lagoon"]
68 --fromList [((fromList ["blue"],fromList ["lagoon"]),2),((fromList ["lagoon"],fromList ["red"]),2)]
69 ----
70 -}
71
72 type Occs = Int
73 type Coocs = Int
74 type Threshold = Int
75
76 removeApax :: Threshold -> Map ([Text], [Text]) Int -> Map ([Text], [Text]) Int
77 removeApax t = DMS.filter (> t)
78
79 cooc :: [[Terms]] -> Map ([Text], [Text]) Int
80 cooc tss = coocOnWithLabel _terms_stem (useLabelPolicy label_policy) tss
81 where
82 terms_occs = occurrencesOn _terms_stem (List.concat tss)
83 label_policy = mkLabelPolicy terms_occs
84
85
86 coocOnWithLabel :: (Ord label, Ord b) => (a -> b) -> (b -> label)
87 -> [[a]] -> Map (label, label) Coocs
88 coocOnWithLabel on' policy tss = mapKeys (delta policy) $ coocOn on' tss
89 where
90 delta :: Arrow a => a b' c' -> a (b', b') (c', c')
91 delta f = f *** f
92
93
94 mkLabelPolicy :: Map Grouped (Map Terms Occs) -> Map Grouped [Text]
95 mkLabelPolicy = DMS.map f where
96 f = _terms_label . fst . maximumWith snd . DMS.toList
97 -- TODO use the Foldable instance of Map instead of building a list
98
99 useLabelPolicy :: Map Grouped [Text] -> Grouped -> [Text]
100 useLabelPolicy m g = case DMS.lookup g m of
101 Just label -> label
102 Nothing -> panic $ "Label of Grouped not found: " <> (pack $ show g)
103 {-
104 labelPolicy :: Map Grouped (Map Terms Occs) -> Grouped -> Label
105 labelPolicy m g = case _terms_label <$> fst <$> maximumWith snd <$> DMS.toList <$> lookup g m of
106 Just label -> label
107 Nothing -> panic $ "Label of Grouped not found: " <> (pack $ show g)
108 -}
109
110 coocOn :: Ord b => (a -> b) -> [[a]] -> Map (b, b) Coocs
111 coocOn f as = DMS.unionsWith (+) $ map (coocOn' f) as
112
113 coocOn' :: Ord b => (a -> b) -> [a] -> Map (b, b) Coocs
114 coocOn' fun ts = DMS.fromListWith (+) xs
115 where
116 ts' = List.nub $ map fun ts
117 xs = [ ((x, y), 1)
118 | x <- ts'
119 , y <- ts'
120 , x >= y
121 ]
122
123 ------------------------------------------------------------------------
124
125 coocOnContexts :: (a -> [Text]) -> [[a]] -> Map ([Text], [Text]) Int
126 coocOnContexts fun = DMS.fromListWith (+) . List.concat . map (coocOnSingleContext fun)
127
128 coocOnSingleContext :: (a -> [Text]) -> [a] -> [(([Text], [Text]), Int)]
129 coocOnSingleContext fun ts = xs
130 where
131 ts' = List.nub $ map fun ts
132 xs = [ ((x, y), 1)
133 | x <- ts'
134 , y <- ts'
135 , x >= y
136 ]
137 ------------------------------------------------------------------------
138
139
140 -- | Compute the grouped occurrences (occ)
141 occurrences :: [Terms] -> Map Grouped (Map Terms Int)
142 occurrences = occurrencesOn _terms_stem
143
144 occurrencesOn :: (Ord a, Ord b) => (a -> b) -> [a] -> Map b (Map a Int)
145 occurrencesOn f = foldl' (\m a -> insertWith (unionWith (+)) (f a) (singleton a 1) m) empty
146
147 -- TODO add groups and filter stops
148
149 sumOcc :: Ord a => [Occ a] -> Occ a
150 sumOcc xs = unionsWith (+) xs
151
152