File size: 2,572 Bytes
24a2e59
be915aa
24a2e59
be915aa
24a2e59
be915aa
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
24a2e59
 
 
be915aa
24a2e59
be915aa
 
 
 
 
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
from transformers import pipeline
import re

class ContextAwareLyricCleaner:
    def __init__(self):
        self.classifier = pipeline("zero-shot-classification", model="facebook/bart-large-mnli")
        self.replacements = {
            r'\bfuck\b': 'frick',
            r'\bshit\b': 'shoot',
            r'\bfucking\b': 'flipping',
            r'\bfucked\b': 'flipped',
            r'\bshitty\b': 'soggy',
            r'\bass\b': 'butt',
            r'\basses\b': 'butts',
            r'\basshole\b': 'jerkface',
            r'\bbitch\b': 'witch',
            r'\bbitches\b': 'witches',
            r'\bdamn\b': 'darn',
            r'\bcunt\b': 'punk',
            r'\bcrap\b': 'junk',
            r'\bdick\b': 'prick',
            r'\bfag\b': 'nerd',
            r'\bfaggot\b': 'loser',
            r'\bmothafucka\b': 'motherlover',
            r'\bmotherfucker\b': 'motherlover',
            r'\bhell\b': 'heck',
            r'\bprick\b': 'jerk',
            r'\bpiss\b': 'pee',
            r'\bpissed\b': 'mad',
            r'\bshithead\b': 'knucklehead',
            r'\bslut\b': 'scout',
            r'\bwhore\b': 'score',
            r'\bwtf\b': 'what the flip',
            r'\bwtf\b': 'what the flip',
            r'\bson of a bitch\b': 'son of a glitch',
            r'\bbastard\b': 'rascal',
            r'\bgod\b': 'gosh',
            r'\blord\b': 'love',
            # Add more...
        }
        self.patterns = {re.compile(k, re.IGNORECASE): v for k, v in self.replacements.items()}
        self.explicit_labels = ["explicit", "offensive", "inappropriate"]
        self.threshold = 0.7  # confidence threshold to consider line explicit

    def is_explicit(self, text: str) -> bool:
        result = self.classifier(text, candidate_labels=self.explicit_labels + ["clean"], multi_label=False)
        scores = dict(zip(result['labels'], result['scores']))
        # Check if any explicit label scores above threshold
        return any(scores.get(label, 0) > self.threshold for label in self.explicit_labels)

    def clean_line(self, line: str) -> str:
        cleaned = line
        for pattern, replacement in self.patterns.items():
            cleaned = pattern.sub(replacement, cleaned)
        return cleaned

    def clean_lyrics(self, lyrics: str) -> str:
        lines = lyrics.split('\n')
        cleaned_lines = []
        for line in lines:
            if self.is_explicit(line):
                cleaned_lines.append(self.clean_line(line))
            else:
                cleaned_lines.append(line)
        return '\n'.join(cleaned_lines)