from smolagents.tools import Tool import pronouncing import json import difflib import string class ParodyWordSuggestionTool(Tool): name = "parody_word_suggester" description = """Suggests rhyming funny words using CMU dictionary pronunciations. Returns similar-sounding words that rhyme, especially focusing on common vowel sounds.""" 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}} output_type = "string" def forward(self, target: str, word_list_str: str, min_similarity: str = "0.5") -> str: """Get rhyming word suggestions.""" import pronouncing import string import json from difflib import SequenceMatcher target = target.lower().strip(string.punctuation) min_similarity = float(min_similarity) suggestions = [] # 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 = pronouncing.phones_for_word(target) if not target_phones: return json.dumps({ "error": f"'{target}' not found in CMU dictionary", "suggestions": [] }, indent=2) target_phones = target_phones[0] target_phone_list = target_phones.split() # Check each word for word in words: word = word.lower().strip(string.punctuation) phones = pronouncing.phones_for_word(word) if phones: word_phones = phones[0] word_phone_list = word_phones.split() # 1. Rhyme score (most important - 60%) rhyme_score = 0 if len(word_phone_list) > 1 and len(target_phone_list) > 1: # Check if words share the same ending (vowel + final consonants) vowel_plus_end = -2 # Index of the vowel in final syllable while vowel_plus_end < -1: if 'A' in word_phone_list[vowel_plus_end] or 'E' in word_phone_list[vowel_plus_end] or 'I' in word_phone_list[vowel_plus_end] or 'O' in word_phone_list[vowel_plus_end] or 'U' in word_phone_list[vowel_plus_end]: break vowel_plus_end += 1 if vowel_plus_end == -1: vowel_plus_end = -2 # Fall back if no vowel found # Check if the ending (from vowel onwards) matches if word_phone_list[vowel_plus_end:] == target_phone_list[vowel_plus_end:]: rhyme_score = 1.0 # 2. Syllable match (25%) target_syl = pronouncing.syllable_count(target_phones) word_syl = pronouncing.syllable_count(word_phones) syllable_score = 1.0 if target_syl == word_syl else 0.0 # 3. Overall similarity (15%) - using string similarity string_similarity = SequenceMatcher(None, target, word).ratio() # Combined score (60% rhyme, 25% syllables, 15% similarity) similarity = (rhyme_score * 0.6) + (syllable_score * 0.25) + (string_similarity * 0.15) if similarity >= min_similarity: suggestions.append({ "word": word, "similarity": round(similarity, 3), "rhyme_match": rhyme_score == 1.0, "syllable_match": syllable_score == 1.0, "string_similarity": round(string_similarity, 3), "syllables": word_syl, "phones": word_phones }) # Sort by similarity score descending suggestions.sort(key=lambda x: x["similarity"], reverse=True) result = { "target": target, "target_syllables": pronouncing.syllable_count(target_phones), "target_phones": target_phones, "suggestions": suggestions } return json.dumps(result, indent=2) def __init__(self, *args, **kwargs): self.is_initialized = False