File size: 15,757 Bytes
8a02978
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
#!/usr/bin/env python3
import argparse
import os
import re
import sys
from dataclasses import dataclass
from typing import List, Sequence, Set, Tuple, Dict, Union, Optional
from hindi_xlit import HindiTransliterator

import torch
from transformers import AutoTokenizer, AutoModelForTokenClassification

BASE_DIR = os.path.dirname(os.path.abspath(__file__))
MODEL_DIR = os.path.join(BASE_DIR, 'hing_bert_module', 'hing-bert-lid')
DICTIONARY_PATH = os.path.join(BASE_DIR, 'hing_bert_module', 'dictionary.txt')
OUTPUT_PATH = os.path.join(BASE_DIR, 'output2.txt')
LOG_INITIALIZED = False

LABEL_MAP = None
LABEL_TO_ID = None

TOKEN_RE = re.compile(r"[A-Za-zĀāĪīŪūṚṛṝḶḷḸḹēēōōṃḥśṣṭḍṇñṅ'’-]+")
COMMON_ENGLISH_STOPWORDS = {
    'a','he', 'an', 'and', 'are', 'as', 'at', 'be', 'because', 'been', 'but', 'by', 'for', 'from',
    'had', 'has', 'have', 'he', 'her', 'here', 'him', 'his', 'how', 'i', 'in', 'is', 'it',
    'its', 'me', 'my', 'no', 'not', 'of', 'on', 'or', 'our', 'she', 'so', 'that', 'the',
    'their', 'them', 'there', 'they', 'this', 'those', 'to', 'was', 'we', 'were', 'what',
    'when', 'where', 'which', 'who', 'whom', 'why', 'will', 'with', 'you', 'your'
}


@dataclass
class TokenPrediction:
    token: str
    label: str
    confidence: float



def load_model(device: str | None = None):
    if device:
        dev = torch.device(device)
    else:
        dev = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
    tokenizer = AutoTokenizer.from_pretrained(MODEL_DIR, local_files_only=True)
    model = AutoModelForTokenClassification.from_pretrained(MODEL_DIR, local_files_only=True)
    model.to(dev)
    model.eval()

    global LABEL_MAP, LABEL_TO_ID
    config = model.config
    if hasattr(config, 'id2label') and config.id2label:
        LABEL_MAP = {int(k): v for k, v in config.id2label.items()}
    else:
        LABEL_MAP = {i: str(i) for i in range(config.num_labels)}
    if hasattr(config, 'label2id') and config.label2id:
        LABEL_TO_ID = {str(k): int(v) for k, v in config.label2id.items()}
    else:
        LABEL_TO_ID = {v: k for k, v in LABEL_MAP.items()}

    return tokenizer, model, dev
def _tokenize(text: str) -> List[str]:
    tokens = [m.group(0) for m in TOKEN_RE.finditer(text)]
    if tokens:
        return tokens
    return text.strip().split()


def _hindi_pattern_score(token: str) -> float:
    t = token.lower()
    if len(t) <= 1:
        return 0.0
    clusters = ['bh', 'chh', 'ch', 'dh', 'gh', 'jh', 'kh', 'ksh', 'ph', 'sh', 'th', 'tr', 'shr', 'str', 'vr', 'kr', 'gy', 'ny', 'arj', 'rj']
    vowels = ['aa', 'ai', 'au', 'ee', 'ii', 'oo', 'ou']
    suffixes = ['a', 'aa', 'am', 'an', 'as', 'aya', 'ana', 'ara', 'iya', 'ika', 'tra']
    score = 0.0
    for c in clusters:
        if c in t:
            score += 0.4
    for v in vowels:
        if v in t:
            score += 0.2
    for suf in suffixes:
        if t.endswith(suf) and len(t) > len(suf):
            score += 0.3
    if t.endswith(('a', 'i', 'o', 'u')):
        score += 0.1
    if re.search(r'[kgcjtdpb]h', t):
        score += 0.2
    return score


def classify_text(
    text: str,
    tokenizer,
    model,
    device,
    threshold: float,
) -> List[TokenPrediction]:
    words = _tokenize(text)
    if not words:
        return []

    batch = tokenizer(
        words,
        return_tensors='pt',
        padding=True,
        truncation=True,
        is_split_into_words=True
    )
    word_ids = batch.word_ids(batch_index=0)
    batch = {k: v.to(device) for k, v in batch.items()}

    with torch.no_grad():
        outputs = model(**batch)
        logits = outputs.logits.squeeze(0)

    word_logits: dict[int, torch.Tensor] = {}
    word_counts: dict[int, int] = {}
    for idx, word_id in enumerate(word_ids):
        if word_id is None:
            continue
        if word_id not in word_logits:
            word_logits[word_id] = logits[idx]
            word_counts[word_id] = 1
        else:
            word_logits[word_id] += logits[idx]
            word_counts[word_id] += 1

    predictions: List[TokenPrediction] = []
    for word_index, word in enumerate(words):
        logits_sum = word_logits.get(word_index)
        if logits_sum is None:
            predictions.append(TokenPrediction(word, 'N/A', 0.0))
            continue
        avg_logits = logits_sum / word_counts[word_index]
        probs = torch.softmax(avg_logits, dim=-1)

        conf, idx = torch.max(probs, dim=-1)
        raw_label = LABEL_MAP.get(int(idx), str(int(idx)))

        hi_idx = LABEL_TO_ID.get('HI') if LABEL_TO_ID else None
        en_idx = LABEL_TO_ID.get('EN') if LABEL_TO_ID else None
        hi_prob = float(probs[hi_idx]) if hi_idx is not None else 0.0
        en_prob = float(probs[en_idx]) if en_idx is not None else float(conf)

        final_label = raw_label
        conf_value = float(conf)

        if hi_idx is not None and hi_prob >= threshold:
            final_label = 'HI'
            conf_value = hi_prob
        elif raw_label == 'HI':
            final_label = 'HI'
            conf_value = hi_prob
        else:
            lower = word.lower()
            pattern_score = _hindi_pattern_score(word)
            is_capitalized = word[:1].isupper() and not word.isupper()

            override = False
            if hi_prob >= threshold - 0.05:
                override = True
            elif hi_prob >= 0.60 and pattern_score >= 0.5:
                override = True
            elif hi_prob >= 0.45 and pattern_score >= 0.6 and is_capitalized:
                override = True
            elif pattern_score >= 0.8 and hi_prob >= 0.40 and lower not in COMMON_ENGLISH_STOPWORDS:
                override = True

            if override and lower not in COMMON_ENGLISH_STOPWORDS:
                final_label = 'HI'
                conf_value = max(hi_prob, threshold - 0.05)
            else:
                final_label = 'EN'
                conf_value = en_prob

        if conf_value < 0.97:
            final_label = 'HI'
            conf_value = max(conf_value, 0.96)

        predictions.append(TokenPrediction(word, final_label, conf_value))

    return predictions


def _print_predictions(predictions: Sequence[TokenPrediction]):
    print("Token\tLabel\tConfidence")
    for pred in predictions:
        print(f"{pred.token}\t{pred.label}\t{pred.confidence:.4f}")


def _init_output_log():
    global LOG_INITIALIZED
    if LOG_INITIALIZED:
        return
    with open(OUTPUT_PATH, 'w', encoding='utf-8') as f:
        f.write('HingBERT-LID session log\n')
    LOG_INITIALIZED = True


def load_dictionary(filename: str = None) -> Dict[str, str]:
    """
    Load the mythology dictionary from file.
    Returns a dictionary mapping English words to Hindi transliterations.
    """
    filename = filename or DICTIONARY_PATH
    dictionary = {}
    try:
        with open(filename, 'r', encoding='utf-8') as f:
            in_dict = False
            for line in f:
                line = line.strip()
                
                # Skip empty lines and comments
                if not line or line.startswith('#'):
                    continue
                    
                # Check if we're in the dictionary section
                if 'MYTHOLOGY_DICTIONARY = {' in line:
                    in_dict = True
                    line = line.split('{', 1)[1].strip()
                    if not line:  # If the line ends after {
                        continue
                
                if not in_dict:
                    continue
                    
                # Process key-value pairs
                if ':' in line:
                    # Handle multi-line entries
                    while not line.rstrip().endswith(','):
                        next_line = next(f, '').strip()
                        if not next_line:
                            break
                        line += ' ' + next_line
                    
                    # Handle the last line which might end with }
                    line = line.split('}')[0].strip()
                    
                    # Split into key-value pairs
                    entries = [e.strip() for e in line.split(',') if ':' in e]
                    
                    for entry in entries:
                        try:
                            key_part, value_part = entry.split(':', 1)
                            key = key_part.strip().strip("'\"")
                            value = value_part.strip().strip("'\"").rstrip('}')
                            if key and value:
                                dictionary[key.lower()] = value
                        except (ValueError, IndexError):
                            continue
                
                # Check for end of dictionary
                if '}' in line and in_dict:
                    break
        
        print(f"✓ Dictionary loaded successfully: {len(dictionary)} words")
        return dictionary
        
    except FileNotFoundError:
        print(f"Warning: Dictionary file '{filename}' not found.")
        print("Proceeding with model-only transliteration.")
        return {}
    except Exception as e:
        print(f"Warning: Error loading dictionary: {str(e)}")
        print("Proceeding with model-only transliteration.")
        return {}


def get_transliteration(word: str, dictionary: Dict[str, str], transliterator, show_source: bool = False) -> Union[str, tuple]:
    """
    Get transliteration for a word.
    First checks dictionary, then falls back to model.
    
    Args:
        word: English word to transliterate
        dictionary: Dictionary mapping English to Hindi
        transliterator: HindiTransliterator instance
        show_source: If True, returns (transliteration, source)
    
    Returns:
        Transliteration string, or tuple (transliteration, source) if show_source=True
    """
    word_lower = word.lower().strip()
    
    # Check dictionary first
    if word_lower in dictionary:
        result = dictionary[word_lower]
        if show_source:
            return result, "dictionary"
        return result
    
    # Fall back to model
    try:
        model_result = transliterator.transliterate(word)
        
        # Handle if model returns a list
        if isinstance(model_result, list):
            result = model_result[0]  # Take first (best) result
        else:
            result = model_result
        
        if show_source:
            return result, "model"
        return result
        
    except Exception as e:
        if show_source:
            return word, "error"
        return word


def _write_predictions(predictions: Sequence[TokenPrediction], source_text: str):
    """Write predictions to output file."""
    global LOG_INITIALIZED
    _init_output_log()
    with open(OUTPUT_PATH, 'a', encoding='utf-8') as f:
        if not LOG_INITIALIZED:
            f.write('\n' + '='*80 + '\n')
            LOG_INITIALIZED = True
        f.write(f'\nSource: {source_text}\n')
        for pred in predictions:
            f.write(f"{pred.token}\t{pred.label}\t{pred.confidence:.4f}\n")


def main():
    parser = argparse.ArgumentParser(description='Test l3cube-pune/hing-bert-lid on text (token-level).')
    parser.add_argument('--device', type=str, default=None, help='torch device (cpu or cuda)')
    parser.add_argument('--text', type=str, default=None, help='Text to classify.')
    parser.add_argument('--threshold', type=float, default=0.80, help='Confidence threshold for Hindi override heuristics (default=0.80)')
    parser.add_argument('--dictionary', type=str, default=DICTIONARY_PATH, help='Path to dictionary file (default: dictionary.txt)')
    args = parser.parse_args()

    tokenizer, model, device = load_model(args.device)
    
    # Load dictionary for transliteration
    dictionary = load_dictionary(args.dictionary)
    transliterator = HindiTransliterator()
    
    hindi_words = set()  # To store unique Hindi words

    if args.text:
        preds = classify_text(args.text, tokenizer, model, device, args.threshold)
        _print_predictions(preds)
        _write_predictions(preds, args.text)
        # Add Hindi words to the set
        hindi_words.update(pred.token for pred in preds if pred.label == 'HI')
        print("\nHindi words found:", ", ".join(hindi_words) if hindi_words else "None")
        return

    print("Interactive mode. Type text lines (QUIT to exit).")
    _init_output_log()
    all_input = []  # Store all input text
    try:
        for line in sys.stdin:
            line = line.rstrip('\n')
            if not line:
                continue
            if line.strip().upper() == 'QUIT':
                break
            all_input.append(line)  # Add to full input
            preds = classify_text(line, tokenizer, model, device, args.threshold)
            _print_predictions(preds)
            _write_predictions(preds, line)
            # Add Hindi words to the set
            hindi_words.update(pred.token for pred in preds if pred.label == 'HI')
            print()
    except (KeyboardInterrupt, EOFError):
        pass
    finally:
        # Print all collected Hindi words before exiting
        if hindi_words:
            print("\nAll Hindi words found:", ", ".join(sorted(hindi_words)))
            
            # Create a mapping of Hindi words to their Devanagari transliterations
            print("\nTransliterated to Devanagari:")
            hindi_to_devanagari = {}
            
            for word in sorted(hindi_words):
                try:
                    devanagari = get_transliteration(word, dictionary, transliterator)
                    hindi_to_devanagari[word] = devanagari
                    source = " (dictionary)" if word.lower() in dictionary else " (model)"
                    print(f"{word} -> {devanagari}{source}")
                except Exception as e:
                    print(f"{word} -> [Error: {str(e)}]")
            
            # Save the output to a file
            output_file = os.path.join(BASE_DIR, 'final_output.txt')
            with open(output_file, 'w', encoding='utf-8') as f:
                # Write original full text
                f.write("=== Original Text ===\n")
                f.write('\n'.join(all_input) + '\n\n')
                
                # Write reconstructed text
                f.write("=== Reconstructed Text with Devanagari ===\n")
                for line in all_input:
                    reconstructed = line
                    for word, devanagari in hindi_to_devanagari.items():
                        # Use a more precise replacement that preserves punctuation and spacing
                        reconstructed = re.sub(
                            rf'(?<![\w\-])({re.escape(word)})(?![\w\-])', 
                            devanagari, 
                            reconstructed,
                            flags=re.IGNORECASE
                        )
                    f.write(reconstructed + '\n')
                
                # Write Hindi words and their transliterations
                f.write("\n=== Hindi Words and Transliterations ===\n")
                for word, devanagari in sorted(hindi_to_devanagari.items()):
                    f.write(f"{word} -> {devanagari}\n")
            
            print(f"\nOutput saved to: {output_file}")
        else:
            print("\nNo Hindi words found.")


if __name__ == '__main__':
    if sys.platform == 'win32':
        try:
            sys.stdout.reconfigure(encoding='utf-8')
        except (AttributeError, TypeError):
            import io
            sys.stdout = io.TextIOWrapper(sys.stdout.buffer, encoding='utf-8', errors='replace')
    main()