sid570 commited on
Commit
2afa51a
·
verified ·
1 Parent(s): 9f7d54c

Update entity_dictionary.py

Browse files
Files changed (1) hide show
  1. entity_dictionary.py +43 -124
entity_dictionary.py CHANGED
@@ -1,12 +1,11 @@
1
  # ============================================================
2
  # ENTITY DICTIONARY
3
- # Used by entity masking API
4
  # ============================================================
5
 
6
  ENTITY_DICTIONARY = {
7
 
8
  # ========================
9
- # 1. DISTRICTS OF MAURITIUS
10
  # ========================
11
  "DISTRICT": [
12
  "Port Louis", "Pamplemousses", "Riviere du Rempart",
@@ -20,80 +19,39 @@ ENTITY_DICTIONARY = {
20
  "CITY": [
21
  "Curepipe", "Vacoas", "Quatre Bornes",
22
  "Beau Bassin", "Rose Hill", "Ebene", "Phoenix",
23
- "Goodlands", "Grand Baie", "Mahebourg", "Flic en Flac",
24
- "Tamarin", "Bel Air", "Roche Bois",
25
- "Piton", "Triolet"
26
  ],
27
 
28
  # ========================
29
  # 3. VILLAGES
30
  # ========================
31
  "VILLAGE": [
32
- "Albion", "Amaury", "Amitie-Gokhoola", "Arsenal",
33
- "Baie du Tombeau", "Baie du Cap", "Bambous",
34
- "Bambous Virieux", "Bananes", "Beau Vallon",
35
- "Bel Air Riviere Seche", "Bel Ombre", "Belle Vue Maurel",
36
- "Benares", "Bois Cheri", "Bois des Amourettes",
37
- "Bon Accueil", "Bramsthan", "Brisee Verdiere",
38
- "Britannia", "Calebasses", "Camp Carol",
39
- "Camp de Masque", "Camp de Masque Pave", "Camp Diable",
40
- "Camp Ithier", "Camp Thorel", "Cap Malheureux",
41
- "Cascavelle", "Case Noyale", "Centre de Flacq",
42
- "Chamarel", "Chamouny", "Chemin Grenier",
43
- "Clemencia", "Cluny", "Congomah", "Cottage",
44
- "Creve Coeur", "D'Epinay", "Dagotier", "Dubreuil",
45
- "Ecroignard", "Esperance", "Esperance Trebuchet",
46
- "Fond du Sac", "Grand Bel Air", "Grand Bois",
47
- "Grand Gaube", "Grand River South East",
48
- "Grand Sable", "Grande Retraite",
49
- "Grande Riviere Noire", "Gros Cailloux",
50
- "L'Avenir", "L'Escalier", "La Flora", "La Gaulette",
51
- "La Laura-Malenga", "Lallmatie", "Laventure",
52
- "Le Hochet", "Le Morne", "Long Mountain",
53
- "Mapou", "Mare d'Albert", "Mare La Chaux",
54
- "Mare Tabac", "Medine", "Melrose", "Midlands",
55
- "Montagne Blanche", "Morcellement St. Andre",
56
- "New Grove", "Notre Dame", "Nouvelle Decouverte",
57
- "Nouvelle France", "Old Grand Port", "Olivia",
58
- "Petit Bel Air", "Petit Raffray",
59
- "Plaine des Papayes", "Plaine des Roches",
60
- "Plaine Magnien", "Pointe aux Piments",
61
- "Poste de Flacq", "Poudre d'Or",
62
- "Poudre d'Or Hamlet", "Providence",
63
- "Quartier Militaire", "Quatre Cocos",
64
- "Quatre Soeurs", "Queen Victoria", "Richelieu",
65
- "Ripailles", "Riviere des Anguilles",
66
- "Riviere des Creoles", "Riviere du Poste",
67
- "Roches Noires", "Roche Terre", "Rose Belle",
68
- "Sebastopol", "Seizieme Mille", "Souillac",
69
- "St Aubin", "St Hubert", "St Julien",
70
- "St Julien D'Hotman", "St Pierre", "Surinam",
71
- "Terre Rouge", "The Vale", "Trois Boutiques",
72
- "Trou aux Biches", "Trou d'Eau Douce",
73
- "Tyack", "Union Park", "Verdun", "Ville Bague"
74
  ],
75
 
76
  # ========================
77
- # 4. LANDMARKS & REGIONS
78
  # ========================
79
  "LANDMARK": [
80
- "Bagatelle", "Caudan Waterfront", "Le Caudan",
81
- "Ebene Cybercity", "Trou aux Biches",
82
- "Mont Choisy", "Chamarel", "Le Morne",
83
- "Blue Bay", "La Cambuse", "Ile aux Cerfs",
84
- "Reduit", "SSR Airport", "Plaisance",
85
- "Cascavelle", "Medine"
86
  ],
87
 
88
  # ========================
89
- # 5. WEEKDAYS (Dictionary entities)
90
  # ========================
91
  "WEEKDAY": [
92
  # Bhojpuri
93
  "Somvar", "Lendi",
94
  "Mangal", "Mangar",
95
- "Budh",
96
- "Guruwar",
97
  "Sukar", "Sukrawar",
98
  "Sani", "Samdi",
99
  "Raviwaar",
@@ -104,68 +62,43 @@ ENTITY_DICTIONARY = {
104
  ],
105
 
106
  # ========================
107
- # 6. MONTHS (Dictionary entities)
108
  # ========================
109
  "MONTH": [
110
- # Bhojpuri / Hindi variants
111
  "January", "Janvari",
112
  "February", "Februari",
113
- "March",
114
- "April",
115
- "May",
116
- "June",
117
- "July",
118
- "August",
119
- "September",
120
- "October",
121
- "November",
122
- "December",
123
-
124
- # Kreol Morisien
125
- "Zanvie",
126
- "Fevriye",
127
- "Mars",
128
- "Avril",
129
- "May",
130
- "Zwin",
131
- "Zilye",
132
- "Aout",
133
- "Septam",
134
- "Oktob",
135
- "Novam",
136
- "Desam"
137
- ],
138
 
 
 
 
 
 
139
 
140
  # ========================
141
- # 7. PERSON NAMES
142
  # ========================
143
  "PERSON": [
144
- "Amit", "Vikash", "Rishi", "Kiran", "Anil", "Sanjay",
145
- "Rakesh", "Nikhil", "Kavita", "Reena", "Meera",
146
- "Neha", "Pooja", "Sunil", "Aisha", "Fatima",
147
- "Farah", "Yusuf", "Imran", "Jean", "Marie",
148
- "Patrick", "Stephan", "Yannick", "Kevin",
149
- "Ramesh", "Champa", "Babita", "Kamal", "Sonu",
150
- "Guddu", "Sidharth", "Sameer", "Sameerchand", "Karuna"
151
  ],
152
 
153
  # ========================
154
- # 8. PRODUCTS / OBJECTS
155
  # ========================
156
  "PRODUCT": [
157
- "Honda Fit", "Honda Vezel", "Suzuki Swift",
158
- "Toyota Aqua", "Toyota Vitz", "Nissan Note",
159
- "Mazda Axela", "iPhone", "Samsung",
160
- "OnePlus", "Pixel", "Redmi"
161
  ],
162
 
163
  # ========================
164
- # 9. REGEX PATTERNS
165
  # ========================
166
  "REGEX_PATTERNS": {
167
  "NUMBER": r"\b\d+(?:,\d{3})*(?:\.\d+)?\b",
168
- "MONEY": r"\b(?:Rs|rs|RS)\s?\d+(?:,\d{3})*\b",
169
  "DATE": r"\b\d{1,2}[\/\-.]\d{1,2}[\/\-.]\d{2,4}\b",
170
  "TIME": r"\b\d{1,2}(?::\d{2})?(?:am|pm|AM|PM)\b",
171
  "PHONE": r"\b[0-9]{8}\b",
@@ -174,11 +107,10 @@ ENTITY_DICTIONARY = {
174
  }
175
 
176
  # ============================================================
177
- # WEEKDAY CANONICAL MAP (used during unmasking)
178
  # ============================================================
179
 
180
  WEEKDAY_MAP = {
181
- # Bhojpuri → Kreol
182
  "somvar": "lindi",
183
  "lendi": "lindi",
184
  "mangal": "mardi",
@@ -191,7 +123,6 @@ WEEKDAY_MAP = {
191
  "samdi": "samdi",
192
  "raviwaar": "dimans",
193
 
194
- # Kreol → Bhojpuri
195
  "lindi": "somvar",
196
  "mardi": "mangal",
197
  "merkredi": "budh",
@@ -202,30 +133,19 @@ WEEKDAY_MAP = {
202
  }
203
 
204
  # ============================================================
205
- # MONTH CANONICAL MAP (used during unmasking)
206
  # ============================================================
207
 
208
- MONTH_MAP = {
209
- # English / Bhojpuri → Kreol
210
- "january": "zanvie",
211
- "janvari": "zanvie",
212
-
213
- "february": "fevriye",
214
- "februari": "fevriye",
215
-
216
- "march": "mars",
217
- "april": "avril",
218
- "may": "may",
219
- "june": "zwin",
220
- "july": "zilye",
221
- "august": "aout",
222
-
223
- "september": "septam",
224
- "october": "oktob",
225
- "november": "novam",
226
- "december": "desam",
227
 
228
- # Kreol → English (canonical)
229
  "zanvie": "january",
230
  "fevriye": "february",
231
  "mars": "march",
@@ -239,4 +159,3 @@ MONTH_MAP = {
239
  "novam": "november",
240
  "desam": "december"
241
  }
242
-
 
1
  # ============================================================
2
  # ENTITY DICTIONARY
 
3
  # ============================================================
4
 
5
  ENTITY_DICTIONARY = {
6
 
7
  # ========================
8
+ # 1. DISTRICTS
9
  # ========================
10
  "DISTRICT": [
11
  "Port Louis", "Pamplemousses", "Riviere du Rempart",
 
19
  "CITY": [
20
  "Curepipe", "Vacoas", "Quatre Bornes",
21
  "Beau Bassin", "Rose Hill", "Ebene", "Phoenix",
22
+ "Goodlands", "Grand Baie", "Mahebourg",
23
+ "Flic en Flac", "Tamarin", "Bel Air",
24
+ "Roche Bois", "Piton", "Triolet"
25
  ],
26
 
27
  # ========================
28
  # 3. VILLAGES
29
  # ========================
30
  "VILLAGE": [
31
+ "Albion", "Amaury", "Arsenal", "Baie du Tombeau",
32
+ "Bambous", "Bel Ombre", "Camp Diable",
33
+ "Grand Gaube", "Lallmatie", "Mapou",
34
+ "New Grove", "Plaine Magnien", "Souillac",
35
+ "Terre Rouge", "Trou aux Biches", "Verdun"
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
36
  ],
37
 
38
  # ========================
39
+ # 4. LANDMARKS
40
  # ========================
41
  "LANDMARK": [
42
+ "Bagatelle", "Caudan Waterfront", "Ebene Cybercity",
43
+ "Le Morne", "Blue Bay", "Ile aux Cerfs",
44
+ "SSR Airport", "Plaisance"
 
 
 
45
  ],
46
 
47
  # ========================
48
+ # 5. WEEKDAYS
49
  # ========================
50
  "WEEKDAY": [
51
  # Bhojpuri
52
  "Somvar", "Lendi",
53
  "Mangal", "Mangar",
54
+ "Budh", "Guruwar",
 
55
  "Sukar", "Sukrawar",
56
  "Sani", "Samdi",
57
  "Raviwaar",
 
62
  ],
63
 
64
  # ========================
65
+ # 6. MONTHS
66
  # ========================
67
  "MONTH": [
68
+ # English / Bhojpuri
69
  "January", "Janvari",
70
  "February", "Februari",
71
+ "March", "April", "May",
72
+ "June", "July", "August",
73
+ "September", "October",
74
+ "November", "December",
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
75
 
76
+ # Kreol
77
+ "Zanvie", "Fevriye", "Mars", "Avril",
78
+ "Me", "Zwin", "Zilye", "Out",
79
+ "Septam", "Oktob", "Novam", "Desam"
80
+ ],
81
 
82
  # ========================
83
+ # 7. PERSON
84
  # ========================
85
  "PERSON": [
86
+ "Amit", "Ramesh", "Sidharth",
87
+ "Sameer", "Karuna", "Jean", "Marie"
 
 
 
 
 
88
  ],
89
 
90
  # ========================
91
+ # 8. PRODUCT
92
  # ========================
93
  "PRODUCT": [
94
+ "Honda Fit", "Toyota Vitz", "iPhone", "Samsung"
 
 
 
95
  ],
96
 
97
  # ========================
98
+ # 9. REGEX
99
  # ========================
100
  "REGEX_PATTERNS": {
101
  "NUMBER": r"\b\d+(?:,\d{3})*(?:\.\d+)?\b",
 
102
  "DATE": r"\b\d{1,2}[\/\-.]\d{1,2}[\/\-.]\d{2,4}\b",
103
  "TIME": r"\b\d{1,2}(?::\d{2})?(?:am|pm|AM|PM)\b",
104
  "PHONE": r"\b[0-9]{8}\b",
 
107
  }
108
 
109
  # ============================================================
110
+ # WEEKDAY MAP (direction-aware)
111
  # ============================================================
112
 
113
  WEEKDAY_MAP = {
 
114
  "somvar": "lindi",
115
  "lendi": "lindi",
116
  "mangal": "mardi",
 
123
  "samdi": "samdi",
124
  "raviwaar": "dimans",
125
 
 
126
  "lindi": "somvar",
127
  "mardi": "mangal",
128
  "merkredi": "budh",
 
133
  }
134
 
135
  # ============================================================
136
+ # MONTH MAPS (Option B)
137
  # ============================================================
138
 
139
+ MONTH_BHO_ENG_TO_KRE = {
140
+ "january": "zanvie", "janvari": "zanvie",
141
+ "february": "fevriye", "februari": "fevriye",
142
+ "march": "mars", "april": "avril", "may": "me",
143
+ "june": "zwin", "july": "zilye", "august": "out",
144
+ "september": "septam", "october": "oktob",
145
+ "november": "novam", "december": "desam"
146
+ }
 
 
 
 
 
 
 
 
 
 
 
147
 
148
+ MONTH_KRE_TO_ENG = {
149
  "zanvie": "january",
150
  "fevriye": "february",
151
  "mars": "march",
 
159
  "novam": "november",
160
  "desam": "december"
161
  }