Upload tool

requirements.txt

@@ -1,2 +1,2 @@
-smolagents
 pronouncing
+smolagents

tool.py

@@ -1,7 +1,8 @@
 from smolagents.tools import Tool
 import pronouncing
-import string
 import json
+import difflib
+import string
 
 class ParodyWordSuggestionTool(Tool):
     name = "parody_word_suggester"
@@ -15,6 +16,7 @@ class ParodyWordSuggestionTool(Tool):
         import pronouncing
         import string
         import json
+        from difflib import SequenceMatcher
     
         target = target.lower().strip(string.punctuation)
         min_similarity = float(min_similarity)
@@ -42,65 +44,49 @@ class ParodyWordSuggestionTool(Tool):
     
         # 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()
             
-                # Calculate full phonetic similarity
-                phonetic_matches = 0
-                max_length = max(len(word_phone_list), len(target_phone_list))
-                min_length = min(len(word_phone_list), len(target_phone_list))
-            
-                # Initial consonant bonus
-                initial_match_score = 0
-                if word_phone_list[0].rstrip('012') == target_phone_list[0].rstrip('012'):
-                    initial_match_score = 1.0  # Reduced bonus for matching first consonant
+                # 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
             
-                # Check if syllable counts match
+                # 2. Syllable match (25%)
                 target_syl = pronouncing.syllable_count(target_phones)
                 word_syl = pronouncing.syllable_count(word_phones)
-                syllable_match = 1.0 if target_syl == word_syl else 0.0
+                syllable_score = 1.0 if target_syl == word_syl else 0.0
             
-                for i in range(max_length):
-                    if i >= min_length:
-                        break
-                    
-                    w_phone = word_phone_list[i]
-                    t_phone = target_phone_list[i]
-                
-                    # Exact phone match
-                    if w_phone == t_phone:
-                        phonetic_matches += 1.0
-                    # Match without stress numbers
-                    elif w_phone.rstrip('012') == t_phone.rstrip('012'):
-                        phonetic_matches += 0.8
-            
-                phonetic_similarity = (phonetic_matches / max_length) + initial_match_score
-            
-                # Rhyme score (focusing on vowel and end consonant)
-                rhyme_score = 0
-                if len(word_phone_list) > 1 and len(target_phone_list) > 1:
-                    # Check final syllable (vowel + final consonant)
-                    if word_phone_list[-1] == target_phone_list[-1]:  # End consonant match
-                        rhyme_score += 1.0
-                    if word_phone_list[-2] == target_phone_list[-2]:  # Vowel match
-                        rhyme_score += 2.0  # Higher weight for vowel match
+                # 3. Overall similarity (15%) - using string similarity
+                string_similarity = SequenceMatcher(None, target, word).ratio()
             
-                # Combined score with new weights prioritizing rhyme
-                # 50% rhyme, 30% syllable match, 20% phonetic similarity
-                similarity = (rhyme_score * 0.5) + (syllable_match * 0.3) + (phonetic_similarity * 0.2)
+                # 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),
-                        "phonetic_similarity": round(phonetic_similarity, 3),
-                        "rhyme_score": round(rhyme_score, 3),
-                        "syllable_match": syllable_match,
-                        "initial_match": initial_match_score > 0,
-                        "syllables": pronouncing.syllable_count(word_phones),
-                        "phones": word_phones,
+                        "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