Upload tool

requirements.txt

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

tool.py

@@ -1,7 +1,7 @@
 from smolagents.tools import Tool
+import string
 import pronouncing
 import json
-import string
 
 class ParodyWordSuggestionTool(Tool):
     name = "parody_word_suggester"
@@ -34,43 +34,35 @@ class ParodyWordSuggestionTool(Tool):
         return None
 
 
-    def _get_phone_similarity(self, phone1: str, phone2: str) -> float:
-        """Calculate similarity between two phones."""
-        # Initialize variables
-        p1 = ""
-        p2 = ""
-        group_str = ""
-        group = []
-    
+    def _phones_are_similar(self, phone1: str, phone2: str) -> bool:
+        """Check if two phones are similar enough to be considered rhyming."""
         # Strip stress markers
         p1 = phone1.rstrip('012')
         p2 = phone2.rstrip('012')
+        group_str = ""
+        group = []
     
         # Exact match
         if p1 == p2:
-            return 1.0
+            return True
         
         # Check similarity groups
         for group_str in self.PHONE_GROUPS.split('|'):
             group = group_str.split(',')
             if p1 in group and p2 in group:
-                return 0.7
+                return True
             
-        # Check broader categories
-        if self._get_phone_type(p1) == self._get_phone_type(p2):
-            return 0.3
-        
-        return 0.0
+        return False
 
 
     def _get_phone_type(self, phone: str) -> str:
         """Get the broad category of a phone."""
         # Strip stress markers
         phone = phone.rstrip('012')
-        v = ""
+        vowel_char = ""
     
         # Vowels
-        if any(v in phone for v in 'AEIOU'):
+        if any(vowel_char in phone for vowel_char in 'AEIOU'):
             return 'vowel'
         
         # Initialize fixed sets for categories
@@ -94,6 +86,51 @@ class ParodyWordSuggestionTool(Tool):
         return 'other'
 
 
+    def _get_rhyme_score(self, phones1: list, phones2: list) -> float:
+        """Calculate rhyme score based on matching phones after primary stressed vowel."""
+        # Initialize variables
+        pos1 = -1
+        pos2 = -1
+        i = 0
+        phone = ""
+        vowel_char = ""
+        rhyme_part1 = []
+        rhyme_part2 = []
+        similarity_count = 0
+        p1 = ""
+        p2 = ""
+    
+        # Find primary stressed vowel position in both words
+        for i, phone in enumerate(phones1):
+            if '1' in phone and any(vowel_char in phone for vowel_char in 'AEIOU'):
+                pos1 = i
+                break
+            
+        for i, phone in enumerate(phones2):
+            if '1' in phone and any(vowel_char in phone for vowel_char in 'AEIOU'):
+                pos2 = i
+                break
+        
+        if pos1 == -1 or pos2 == -1:
+            return 0.0
+    
+        # Get all phones after and including the stressed vowel
+        rhyme_part1 = phones1[pos1:]
+        rhyme_part2 = phones2[pos2:]
+    
+        # Check if lengths match
+        if len(rhyme_part1) != len(rhyme_part2):
+            return 0.0
+    
+        # Calculate similarity score for rhyming part
+        for p1, p2 in zip(rhyme_part1, rhyme_part2):
+            if self._phones_are_similar(p1, p2):
+                similarity_count += 1
+    
+        # Return score based on how many phones were similar
+        return similarity_count / len(rhyme_part1) if rhyme_part1 else 0.0
+
+
     def _calculate_phone_sequence_similarity(self, phones1: list, phones2: list) -> float:
         """Calculate similarity based on matching phones in sequence."""
         if not phones1 or not phones2:
@@ -103,7 +140,6 @@ class ParodyWordSuggestionTool(Tool):
         total_similarity = 0.0
         i = 0
         similarity = 0.0
-    
         comparisons = max(len(phones1), len(phones2))
     
         # Compare each position
@@ -114,6 +150,31 @@ class ParodyWordSuggestionTool(Tool):
         return total_similarity / comparisons if comparisons > 0 else 0.0
 
 
+    def _get_phone_similarity(self, phone1: str, phone2: str) -> float:
+        """Calculate similarity between two phones."""
+        # Initialize variables
+        p1 = phone1.rstrip('012')
+        p2 = phone2.rstrip('012')
+        group_str = ""
+        group = []
+    
+        # Exact match
+        if p1 == p2:
+            return 1.0
+        
+        # Check similarity groups
+        for group_str in self.PHONE_GROUPS.split('|'):
+            group = group_str.split(',')
+            if p1 in group and p2 in group:
+                return 0.7
+            
+        # Check broader categories
+        if self._get_phone_type(p1) == self._get_phone_type(p2):
+            return 0.3
+        
+        return 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))
@@ -123,24 +184,35 @@ class ParodyWordSuggestionTool(Tool):
 
     def _calculate_similarity(self, word1, phones1, word2, phones2):
         """Calculate similarity based on multiple factors."""
-        # Initialize phone lists and scores
+        # Initialize variables
         phone_list1 = phones1.split()
         phone_list2 = phones2.split()
         rhyme_score = 0.0
         phone_sequence_score = 0.0
         length_score = 0.0
-        vowel1 = None
-        vowel2 = None
+        similarity = 0.0
     
-        # 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
+        # Get rhyme score using new method
+        rhyme_score = self._get_rhyme_score(phone_list1, phone_list2)
     
-        # 2. Phone sequence similarity (30%)
-        phone_sequence_score = self._calculate_phone_sequence_similarity(phone_list1, phone_list2)
+        # If rhyme score is too low (e.g. below 0.8), consider it a non-rhyme
+        if rhyme_score < 0.8:
+            return {
+                "similarity": 0.0,
+                "rhyme_score": 0.0,
+                "phone_sequence_score": 0.0,
+                "length_score": 0.0,
+                "details": {
+                    "primary_vowel1": self._get_primary_vowel(phone_list1),
+                    "primary_vowel2": self._get_primary_vowel(phone_list2),
+                    "phone_count1": len(phone_list1),
+                    "phone_count2": len(phone_list2),
+                    "matching_phones": 0
+                }
+            }
     
-        # 3. Length similarity (20%)
+        # Calculate other scores only if words rhyme closely enough
+        phone_sequence_score = self._calculate_phone_sequence_similarity(phone_list1, phone_list2)
         length_score = self._calculate_length_similarity(phone_list1, phone_list2)
     
         # Combined weighted score
@@ -156,8 +228,8 @@ class ParodyWordSuggestionTool(Tool):
             "phone_sequence_score": round(phone_sequence_score, 3),
             "length_score": round(length_score, 3),
             "details": {
-                "primary_vowel1": vowel1,
-                "primary_vowel2": vowel2,
+                "primary_vowel1": self._get_primary_vowel(phone_list1),
+                "primary_vowel2": self._get_primary_vowel(phone_list2),
                 "phone_count1": len(phone_list1),
                 "phone_count2": len(phone_list2),
                 "matching_phones": round(phone_sequence_score * len(phone_list1))