Hugging Face's logo

Spaces:
Warrior786
/
memoni-labeling
Sleeping

App Files Community
memoni-labeling
File size: 4,555 Bytes
c9f8865
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
0e8b341
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
"""
Memoni Roman/Urdu normalization rules.
Used by:
  - labeling app (sidebar reference for volunteers)
  - Module 3.5 pipeline (normalize pseudo-labeled text before mT5 training)

Rule: one canonical Roman form per word.
Urdu script form is the equivalent for volunteers who prefer to write in Urdu.
"""

# Roman variant → canonical Roman
# All keys lowercase. Apply after lowercasing input.
CANONICAL = {
    # I / in
    'main': 'mein', 'me': 'mein', 'mn': 'mein', 'mien': 'mein',

    # is
    'hy': 'hai', 'he': 'hai', 'hay': 'hai',

    # are
    'hen': 'hain', 'hein': 'hain',

    # and
    'or': 'aur', 'ur': 'aur', 'oor': 'aur',

    # no / not
    'nahin': 'nahi', 'nai': 'nahi', 'ni': 'nahi', 'nae': 'nahi', 'nhi': 'nahi',

    # what / did
    'kia': 'kya', 'kiya': 'kya', 'ke': 'kya',

    # come
    'aavo': 'aao', 'aaw': 'aao', 'aawo': 'aao',

    # go
    'jaavo': 'jao', 'jaw': 'jao', 'jawo': 'jao',

    # this
    'ye': 'yeh', 'ya': 'yeh', 'yae': 'yeh',

    # that
    'wo': 'woh', 'wu': 'woh',

    # house
    'gher': 'ghar', 'ghur': 'ghar',

    # say / said
    'keha': 'kaha', 'kehya': 'kaha',

    # person
    'bunda': 'banda', 'bende': 'banda',

    # now
    'ub': 'ab', 'abb': 'ab',

    # very
    'buhat': 'bohat', 'bhut': 'bohat', 'bahut': 'bohat',

    # good
    'accha': 'acha', 'atcha': 'acha', 'achha': 'acha',

    # yes
    'haan': 'han', 'hun': 'han',

    # today
    'aj': 'aaj',

    # come (imperative, singular)
    'awo': 'ao', 'aavo': 'ao',

    # with
    'saath': 'sath', 'saath': 'sath',

    # here
    'idhar': 'ider', 'udhar': 'uder',

    # how
    'kaise': 'kese', 'kaisa': 'kesa',

    # why
    'kyun': 'kyun', 'kiun': 'kyun',

    # where
    'kahan': 'kahan', 'kahan': 'kahan',

    # when
    'kab': 'kab',

    # want
    'chahiye': 'chahye', 'chahie': 'chahye',

    # money
    'paisa': 'pesa', 'paesa': 'pesa',

    # brother
    'bhai': 'bhai', 'bhi': 'bhai',

    # sister
    'bhen': 'bhen', 'behan': 'bhen',
}


# Reference table shown to volunteers in the labeling UI
# Format: (canonical_roman, urdu_script, english_meaning)
REFERENCE_TABLE = [
    ('mein',  'میں',  'I / in'),
    ('hai',   'ہے',   'is'),
    ('hain',  'ہیں',  'are'),
    ('aur',   'اور',  'and'),
    ('nahi',  'نہیں', 'no / not'),
    ('kya',   'کیا',  'what / did'),
    ('aao',   'آؤ',   'come (plural)'),
    ('ao',    'آو',   'come (singular)'),
    ('jao',   'جاؤ',  'go'),
    ('yeh',   'یہ',   'this'),
    ('woh',   'وہ',   'that'),
    ('ghar',  'گھر',  'house'),
    ('bohat', 'بہت',  'very'),
    ('acha',  'اچھا', 'good / okay'),
    ('han',   'ہاں',  'yes'),
    ('na',    'نہ',   'no (soft)'),
    ('aaj',   'آج',   'today'),
    ('kab',   'کب',   'when'),
    ('kahan', 'کہاں', 'where'),
    ('kyun',  'کیوں', 'why'),
    ('kese',  'کیسے', 'how'),
    ('sath',  'ساتھ', 'with'),
    ('ab',    'اب',   'now'),
    ('bhai',  'بھائی','brother'),
    ('bhen',  'بہن',  'sister'),
    ('kaha',  'کہا',  'said'),
    ('banda', 'بندہ', 'person / guy'),
    ('pesa',  'پیسہ', 'money'),
]


def normalize(text: str) -> str:
    """Normalize a Roman Memoni transcript to canonical forms."""
    words = text.lower().split()
    return ' '.join(CANONICAL.get(w, w) for w in words)


# ── Urdu script → Roman transliteration ──────────────────────────────────────
_U2R = {
    'ا': 'a',  'آ': 'aa', 'ب': 'b',  'پ': 'p',  'ت': 't',  'ٹ': 't',
    'ث': 's',  'ج': 'j',  'چ': 'ch', 'ح': 'h',  'خ': 'kh', 'د': 'd',
    'ڈ': 'd',  'ذ': 'z',  'ر': 'r',  'ڑ': 'r',  'ز': 'z',  'ژ': 'zh',
    'س': 's',  'ش': 'sh', 'ص': 's',  'ض': 'z',  'ط': 't',  'ظ': 'z',
    'ع': '',   'غ': 'gh', 'ف': 'f',  'ق': 'q',  'ک': 'k',  'گ': 'g',
    'ل': 'l',  'م': 'm',  'ن': 'n',  'ں': 'n',  'و': 'w',  'ہ': 'h',
    'ھ': 'h',  'ی': 'y',  'ے': 'e',  'ء': '',   'ئ': 'y',  'ؤ': 'w',
    'ة': 't',  'إ': 'i',  'أ': 'a',  'ٰ': '',   '،': ',',  '؟': '?',
    # diacritics
    'َ': 'a',  'ِ': 'i',  'ُ': 'u',  'ّ': '',   'ْ': '',   'ً': 'an',
    'ٍ': 'in', 'ٌ': 'un',
    ' ': ' ',
}

def urdu_to_roman(text: str) -> str:
    """Best-effort character-level transliteration of Urdu script to Roman."""
    out = []
    for ch in text:
        out.append(_U2R.get(ch, ch if ord(ch) < 128 else ''))
    result = ''.join(out).strip()
    # Collapse multiple spaces
    import re
    return re.sub(r' +', ' ', result)