Spaces:
Configuration error
Configuration error
| from smolagents.tools import Tool | |
| import pronouncing | |
| import json | |
| import string | |
| class ParodyWordSuggestionTool(Tool): | |
| name = "parody_word_suggester" | |
| description = "Suggests rhyming funny words using CMU dictionary pronunciations." | |
| inputs = {'target': {'type': 'string', 'description': 'The word you want to find rhyming alternatives for'}, 'word_list_str': {'type': 'string', 'description': 'JSON string of word list (e.g. \'["word1", "word2"]\')'}, 'min_similarity': {'type': 'string', 'description': 'Minimum similarity threshold (0.0-1.0)', 'nullable': True, 'default': '0.5'}, 'custom_phones': {'type': 'object', 'description': 'Optional dictionary of custom word pronunciations', 'nullable': True, 'default': None}} | |
| output_type = "string" | |
| RHYME_WEIGHT = 0.5 | |
| PHONE_SEQUENCE_WEIGHT = 0.3 | |
| LENGTH_WEIGHT = 0.2 | |
| def _get_word_phones(self, word, custom_phones=None): | |
| """Get phones for a word, checking custom dictionary first.""" | |
| if custom_phones and word in custom_phones: | |
| return custom_phones[word]["primary_phones"] | |
| import pronouncing | |
| phones = pronouncing.phones_for_word(word) | |
| return phones[0] if phones else None | |
| def _get_primary_vowel(self, phones: list) -> str: | |
| """Get the primary stressed vowel from phone list.""" | |
| v = "" | |
| for phone in phones: | |
| if '1' in phone and any(v in phone for v in 'AEIOU'): | |
| return phone.rstrip('012') | |
| return None | |
| def _calculate_phone_sequence_similarity(self, phones1: list, phones2: list) -> float: | |
| """Calculate similarity based on matching phones in sequence.""" | |
| p = "" | |
| if not phones1 or not phones2: | |
| return 0.0 | |
| # Strip stress markers for comparison | |
| clean_phones1 = [p.rstrip('012') for p in phones1] | |
| clean_phones2 = [p.rstrip('012') for p in phones2] | |
| matches = 0 | |
| total_comparisons = max(len(clean_phones1), len(clean_phones2)) | |
| # Compare phones in sequence | |
| for i in range(min(len(clean_phones1), len(clean_phones2))): | |
| if clean_phones1[i] == clean_phones2[i]: | |
| matches += 1 | |
| return matches / total_comparisons if total_comparisons > 0 else 0.0 | |
| def _calculate_length_similarity(self, phones1: list, phones2: list) -> float: | |
| """Calculate similarity based on phone length.""" | |
| max_length = max(len(phones1), len(phones2)) | |
| length_diff = abs(len(phones1) - len(phones2)) | |
| return 1.0 - (length_diff / max_length) if max_length > 0 else 0.0 | |
| def _calculate_similarity(self, word1, phones1, word2, phones2): | |
| """Calculate similarity based on multiple factors.""" | |
| # Initialize phone lists | |
| phone_list1 = phones1.split() | |
| phone_list2 = phones2.split() | |
| # 1. Rhyme score (50%) - based on primary vowel | |
| vowel1 = self._get_primary_vowel(phone_list1) | |
| vowel2 = self._get_primary_vowel(phone_list2) | |
| rhyme_score = 1.0 if vowel1 and vowel2 and vowel1 == vowel2 else 0.0 | |
| # 2. Phone sequence similarity (30%) | |
| phone_sequence_score = self._calculate_phone_sequence_similarity(phone_list1, phone_list2) | |
| # 3. Length similarity (20%) | |
| length_score = self._calculate_length_similarity(phone_list1, phone_list2) | |
| # Combined weighted score | |
| similarity = ( | |
| (rhyme_score * self.RHYME_WEIGHT) + | |
| (phone_sequence_score * self.PHONE_SEQUENCE_WEIGHT) + | |
| (length_score * self.LENGTH_WEIGHT) | |
| ) | |
| return { | |
| "similarity": round(similarity, 3), | |
| "rhyme_score": round(rhyme_score, 3), | |
| "phone_sequence_score": round(phone_sequence_score, 3), | |
| "length_score": round(length_score, 3), | |
| "details": { | |
| "primary_vowel1": vowel1, | |
| "primary_vowel2": vowel2, | |
| "phone_count1": len(phone_list1), | |
| "phone_count2": len(phone_list2), | |
| "matching_phones": round(phone_sequence_score * len(phone_list1)) | |
| } | |
| } | |
| def forward(self, target: str, word_list_str: str, min_similarity: str = "0.5", custom_phones: dict = None) -> str: | |
| import pronouncing | |
| import string | |
| import json | |
| # Initialize variables | |
| target = target.lower().strip(string.punctuation) | |
| min_similarity = float(min_similarity) | |
| suggestions = [] | |
| valid_words = [] | |
| invalid_words = [] | |
| # Parse JSON string to list | |
| try: | |
| words = json.loads(word_list_str) | |
| except json.JSONDecodeError: | |
| return json.dumps({ | |
| "error": "Invalid JSON string for word_list_str", | |
| "suggestions": [] | |
| }, indent=2) | |
| # Get target pronunciation | |
| target_phones = self._get_word_phones(target, custom_phones) | |
| if not target_phones: | |
| return json.dumps({ | |
| "error": f"Target word '{target}' not found in dictionary or custom phones", | |
| "suggestions": [] | |
| }, indent=2) | |
| # Filter word list | |
| for word in words: | |
| word = word.lower().strip(string.punctuation) | |
| if self._get_word_phones(word, custom_phones): | |
| valid_words.append(word) | |
| else: | |
| invalid_words.append(word) | |
| if not valid_words: | |
| return json.dumps({ | |
| "error": "No valid words found in dictionary or custom phones", | |
| "invalid_words": invalid_words, | |
| "suggestions": [] | |
| }, indent=2) | |
| # Check each word | |
| for word in valid_words: | |
| word_phones = self._get_word_phones(word, custom_phones) | |
| if word_phones: | |
| similarity_result = self._calculate_similarity(word, word_phones, target, target_phones) | |
| if similarity_result["similarity"] >= min_similarity: | |
| suggestions.append({ | |
| "word": word, | |
| "similarity": similarity_result["similarity"], | |
| "rhyme_score": similarity_result["rhyme_score"], | |
| "phone_sequence_score": similarity_result["phone_sequence_score"], | |
| "length_score": similarity_result["length_score"], | |
| "phones": word_phones, | |
| "is_custom": word in custom_phones if custom_phones else False, | |
| "details": similarity_result["details"] | |
| }) | |
| # Sort by similarity score descending | |
| suggestions.sort(key=lambda x: x["similarity"], reverse=True) | |
| result = { | |
| "target": target, | |
| "target_phones": target_phones, | |
| "invalid_words": invalid_words, | |
| "suggestions": suggestions | |
| } | |
| return json.dumps(result, indent=2) | |
| def __init__(self, *args, **kwargs): | |
| self.is_initialized = False | |